Encyclopedia of Nanoscience and Nanotechnology 9781588830586, 1588830586

v. 1. A-Ch -- v. 2. Cl-E -- v. 3. El-H -- v. 4. Hy-M -- v. 5. Mag-Mu -- v. 6. Nano A-M -- v. 7. Nano Me-T -- v. 8. Ne-P

469 98 24MB

English Pages 921 Year 2004

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
ToC......Page 1
p001......Page 4
p027......Page 30
p051......Page 54
p105......Page 108
p133......Page 136
p153......Page 156
p171......Page 174
p187......Page 190
p199......Page 202
p221......Page 224
p237......Page 240
p283......Page 286
p305......Page 308
p317......Page 320
p327......Page 330
p337......Page 340
p371......Page 374
p389......Page 392
p399......Page 402
p429......Page 432
p443......Page 446
p475......Page 478
p495......Page 498
p505......Page 508
p537......Page 540
p565......Page 568
p587......Page 590
p627......Page 630
p639......Page 642
p695......Page 698
p717......Page 720
p731......Page 734
p745......Page 748
p755......Page 758
p773......Page 776
p793......Page 796
p819......Page 822
p837......Page 840
p857......Page 860
p895......Page 898
Recommend Papers

Encyclopedia of Nanoscience and Nanotechnology
 9781588830586, 1588830586

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Ingenta Select: Electronic Publishing Solutions

1 of 3

file:///C:/Vtemp/encyclopediann02/02/02.files/m_cp1-1.htm

Encyclopedia of Nanoscience and Nanotechnology Volume 2 Number 1 2004 Cluster-Assembled Nanostructured Carbon

1

P. Milani; A. Podestà; E. Barborini Colloidal Gold

27

Victor H. Perez-Luna; Kadir Aslan; Pravin Betala Computational Atomic Nanodesign

51

Michael Rieth; Wolfram Schommers Conducting Polymer Nanostructures

105

Rupali Gangopadhyay Conducting Polymeric Nanomaterials

133

Silmara Neves; Wilson A. Gazotti; Marco-A. De Paoli Conducting Polymer Nanotubes

153

Meixiang Wan Confined Molecules in Nanopores

171

H. Tanaka; M. El-Merraoui; H. Kanoh; K. Kaneko Copper Chloride Quantum Dots

187

Keiichi Edamatsu; Tadashi Itoh Core-Shell Nanoparticles

199

Guido Kickelbick; Luis M. Liz-Marzán Crystal Engineering of Metallic Nanoparticles

221

A. Hernández Creus; Y. Gimeno; R. C. Salvarezza; A. J. Arvia Crystallography and Shape of Nanoparticles and Clusters

237

J. M. Montejano-Carrizales; J. L. Rodríguez-López; C. Gutierrez-Wing; M. Miki-Yoshida; M. José-Yacaman Cyclodextrins and Molecular Encapsulation

283

J. Szejtli Deformation Behavior of Nanocrystalline Materials

305

Alla V. Sergueeva; Amiya K. Mukherjee Dendrimer-Metal Nanocomposites

317

Kunio Esumi Dendritic Encapsulation

327

Christopher S. Cameron; Christopher B. Gorman

4/21/2007 8:55 PM

Ingenta Select: Electronic Publishing Solutions

2 of 3

file:///C:/Vtemp/encyclopediann02/02/02.files/m_cp1-1.htm

Diamond Nanocrystals

337

Tarun Sharda; Somnath Bhattacharyya Dielectric Properties of Nanoparticles

371

S. Berger Diffusion and Thermal Vacancy Formation in Nanocrystalline Materials

389

U. Brossmann; R. Würschum; H.-E. Schaefer Diffusion in Nanomaterials

399

Yu. Kaganovskii; L. N. Paritskaya Dip-Pen Nanolithography

429

Benjamin W. Maynor; Jie Liu Disulfide-Bond Associated Protein Folding

443

Hideki Tachibana; Shin-ichi Segawa DNA Electronics

475

Massimiliano Di Ventra; Michael Zwolak DNA-Based Nanodevices

495

Friedrich C. Simmel; Bernard Yurke Doped Carbon Nanotubes

505

Rui-Hua Xie; Jijun Zhao; Qin Rao Doped II-VI Semiconductor Nanoparticles

537

S. K. Kulkarni Double- and Triple-Decker Phthalocyanines/Porphyrins

565

Tatyana N. Lomova; Marija E. Klyueva Doubly Connected Mesoscopic Superconductors

587

B. J. Baelus; F. M. Peeters Drug Nanocrystals of Poorly Soluble Drugs

627

Rainer H. Müller; Aslihan Akkar Dye/Inorganic Nanocomposites

639

A. Lerf; P. apková Dynamic Processes in Magnetic Nanostructures

695

B. Hillebrands; S. O. Demokritov Dynamics of Metal Nanoclusters

717

Paul-Gerhard Reinhard; Eric Suraud Elastic Models for Carbon Nanotubes

731

C. Q. Ru

4/21/2007 8:55 PM

Ingenta Select: Electronic Publishing Solutions

3 of 3

file:///C:/Vtemp/encyclopediann02/02/02.files/m_cp1-1.htm

Electrically Detected Magnetic Resonance

745

Carlos F. O. Graeff Electrochemical Fabrication of Metal Nanowires

755

Huixin He; Nongjian J. Tao Electrochemical Insertion of Lithium in Carbon Nanotubes

773

G. Maurin; F. Henn Electrochemical Nanoelectrodes

793

Irina Kleps Electrochemical Self-Assembly of Oxide/Dye Composites

819

Tsukasa Yoshida; Derck Schlettwein Electrochemical Synthesis of Semiconductor Nanowires

837

Dongsheng Xu; Guolin Guo Electrochemistry of Silicate-Based Nanomaterials

857

Alain Walcarius Electrodeposited Nanogranular Magnetic Cobalt Alloys

895

Valery Fedosyuk Copyright © 2004 American Scientific Publishers

4/21/2007 8:55 PM

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Cluster-Assembled Nanostructured Carbon P. Milani, A. Podestà, E. Barborini Università degli Studi di Milano, Milan, Italy

CONTENTS 1. Introduction 2. Gas-Phase Synthesis of Carbon Clusters 3. Carbon Cluster Beam Sources 4. Control of Film Nanostructure by Cluster Selection 5. Structural and Functional Characterization 6. Properties and Applications 7. Applications of Cluster-Assembled Carbon Glossary References

1. INTRODUCTION Since the discovery of fullerenes and carbon nanotubes and the recent progress in the synthesis of diamond-like nanostructures, carbon has occupied a strategic position in materials science and technology as one of the most versatile and far-reaching materials [1–3]. An arsenal of advanced growth methods now has the potential to provide a large variety of novel carbon materials with tailored properties and functions. In particular, the recent development of cluster assembling techniques has greatly contributed to the notion and practice of nanostructured materials [4]. Nanostructured carbon, in particular, offers a growing number of applications, which apparently depend on its nanoscale constitution. In order to establish a link between nanostructure and materials performances, characterization and manipulation techniques have to be developed and fully mastered on the nanometric scale. In particular, the structural and functional properties of carbon critically depend on the ratio between the numbers of sp (carbyne-like), sp2 (graphite-like), and sp3 (diamond-like) bonds. The control of such a ratio, which has become possible through the most recent growth techniques such as, for example, the one based on supersonic cluster beam assembling [4–6], enables one to synthesize a variety of carbon thin films of great interest in tribology ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

(self-lubrication, wear-resistant, superhard coatings), electronics (field emission for flat-panel displays), or electrochemistry (molecular sieves, ionic and molecular insertion for various applications, including supercapacitors) [5]. Carbon particles and soot attracted the attention of the scientific community well before the discovery of fullerenes [7]. The synthesis and characterization of soot are of particular relevance in many fields of pure and applied science such as astrophysics, electrochemistry, and catalysis [5, 7]. Hundreds of different carbon soots generated by arc discharge, pyrolisis, or sooting flames have been studied since the 1960s [7]. The unifying point of all of the different approaches is the nanoscale structure of these materials, where one hopes to recognize universal building blocks and to determine assembling strategies for the production of materials with tailored properties. The structural and functional properties of carbon thin films assembled atom by atom and, in particular, the sp2 /sp3 ratio, are largely determined by the kinetic energy of the particles impinging on the substrate during the film growth [3]. The use of clusters instead of atoms, as building blocks, can open new possibilities for the synthesis of materials where the structural and functional properties are also determined by the hierarchical organization of units with dimensions ranging from the nanoscopic to the mesoscopic scale [4, 5]. Experimental and theoretical investigations have shown that small carbon clusters (roughly below 40 atoms) have chain or ring structures, whereas larger clusters have the tendency to form three-dimensional cage-like structures characterized by sp2 coordination [8–17]. The low-energy deposition of different carbon clusters preserving the original structure thus would be a very powerful technique for the creation of materials with controlled hybridization, and hence controlled physicochemical properties [4]. For example, films grown using a cluster mass distribution depleted by the large fullerene-like clusters are expected to have a very disordered compact structure; on the other hand, films grown with a cluster distribution rich in the large fullerene-like component should be characterized by a disorder graphitic structure with a large porosity [18, 19].

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (1–26)

Cluster-Assembled Nanostructured Carbon

2

2. GAS-PHASE SYNTHESIS OF CARBON CLUSTERS The deposition process of nanostructured carbon films starting from clusters produced in the gas phase can be divided into the following steps: 1. 2. 3. 4.

cluster cluster cluster cluster

formation extraction selection deposition.

Gas-phase synthesis of clusters is a very efficient method based on the growth of particles from a monomeric vapor [4, 20, 21]. Typically, a Knudsen cell for the production of atomic beams of thermally vaporizable substances can be used for the generation of cluster beams. However, the presence of clusters in a standard effusive beam as produced by a thermal evaporation source is residual, and therefore a gas-aggregation process should be activated by nonequilibrium cooling [20]. This consists of mixing in a condensation cell the atomic vapor with a cold buffer gas (usually a noble gas). The clusters are then extracted by letting the mixture expand through a nozzle. The geometry and dimensions of the source influence the residence time of the clusters before expansion, and hence the cluster-size distribution [4]. Generally speaking, the requisites that one should look at for assembling nanoparticles are the control of mass distribution, structure, and chemical reactivity. Moreover, one should be able to control the degree of coalescence of the clusters during the formation of the nanostructured materials [4]. The production of clusters of refractory materials like carbon requires an efficient vaporization method, such as laser ablation, arc discharge, sputtering, and so on [4, 22, 23]. These methods must be coupled with an efficient cluster extraction, which can be obtained by a gas stream. Cluster mass selection is also very important in order to obtain films and materials with tailored properties. This can be achieved by varying the particle condensation conditions prior to extraction; however, only a coarse control of mass distribution is possible with this method [4, 20]. Clusters can also be selected, once extracted from the source, by mechanical or electromagnetic filtering methods [20, 21]. These have the disadvantage of considerably lowering the cluster flux, and thus being impractical for many applications where mass production is required.

Figure 1 shows a schematic representation of the experimental configuration of the so-called anodic jet carbon arc (AJCA) deposition technique used to deposit nanocomposite carbon films. Nitrogen is injected into the chamber, and is directed toward the arcing region through an orifice at the side of a specially machined hollow anode. During the discharge, the ionization of nitrogen takes place [5, 22, 24]. The working pressure in the chamber is kept below 10−3 mbar. The electric arc is driven by an ac voltage of 22–24 V, with the arc current typically between 70 and 100 A. The use of an ac voltage is not very common, and effectively results in each electrode oscillating between being an anode and a cathode [5]. The production of cluster-assembled carbon foams has been reported by using high-repetition-rate laser ablation in inert gas atmosphere [25, 26]. A silicon substrate placed at a distance of 17 cm was used to collect the particles ablated from a graphite target [27]. This form of cluster-assembled carbon is characterized by a coexistence of sp2 and sp3 bonds and a very low density similar to that of carbon aerogels [28]. Pulsed laser deposition with an excimer laser (KrF) has also been used to produce a form of nanoporous carbon films. By varying the pressure in the deposition chamber, control of density and porosity has been demonstrated [29]. The above-mentioned approaches do not have control of cluster mass distribution produced during the ablation process or of the kinetic energies of the nanoparticles.

2.2. Supersonic Cluster Beams Supersonic cluster beam deposition (SCBD) allows a reasonably precise control of cluster mass distribution and kinetic energy, with the possibility of obtaining very large deposition rates [4, 24]. The availability of intense and stable cluster sources makes cluster beam deposition a viable technique for the synthesis of cluster-assembled carbon films [30, 31]. A supersonic beam is schematically described as a gas stream expanding very rapidly from a high-pressure region

Graphite Cathode Arc spot

plasma

2.1. Sputtering and Laser Ablation The production of carbon-based systems having a structure at the nanoscale due to the incorporation of clusters can be obtained using arc-discharge methods [22, 24]. Carbon films with inclusions of fullerene-like structures and nanotube fragments have been produced with a filtered cathodic arc apparatus, where the arc occurs in a localized highpressure (10 torr) region. Due to the presence of the inert gas, condensation takes place, and a beam of monomer ions and clusters is formed and deposited. The incorporation of three-dimensional networks, consisting of buckled planes with fullerene-like features, induces the formation of films with very high values of hardness and elasticity [5, 22, 24].

Graphite anode

Gas stream To pump

Si substrate

Carbon species

Figure 1. Schematic representation of the experimental configuration of the so-called anodic jet carbon arc (AJCA) deposition technique. Reprinted with permission from [5], G. A. J. Amaratunga et al., in “Nanostructured Carbon for Advanced Applications.” Kluwer, Dordrecht, 2001. © 2001, Kluwer Academic.

Cluster-Assembled Nanostructured Carbon

(source), through a nozzle, to a low-pressure region [32]. The characteristics of the beam are mainly determined by the size and shape of the nozzle and by the pressure difference between the two regions. Compared to effusive beams used in molecular beam epitaxy, supersonic beams provide higher intensity and directionality, allowing the deposition of films with very high growth rates [32]. When a heavy species is diluted in a lighter one (as in the case of clusters diluted in He or another inert gas) and the mixture is expanded, there is formation of a seeded supersonic beam. Seeded supersonic beams are extensively used to cool and to accelerate heavy species such as clusters, and they are particularly attractive for cluster beam deposition [4, 33, 34]. A supersonic expansion can be obtained by imposing a pressure ratio less than Pb /P0 = 0478 across a convergent nozzle driving an isentropic flow expansion, where P0 and Pb are the stagnation and the background pressures, respectively. The expansion through a convergent nozzle will always take place in a subsonic regime regardless of the amount of the applied pressure ratio. Outside the converging nozzle, depending on the pressure ratio, the flow will supersonically expand to pressures even much lower than the background. A normal shock, known as the Mach disk, matches the pressure inside the jet to the background, and closes an area called the zone of silence. The location of the Mach disk has been empirically determined as being only a function of the pressure ratio, and independent of the fluid nature and nozzle geometry [32]. In general, an isentropic supersonic jet led by a shockfree converging–diverging nozzle is advantageous compared to a nonisentropic underexpanded jet obtained by a convergent nozzle. However, depending on the Reynolds number, a converging–diverging nozzle requires a slow increase in cross-sectional area to avoid boundary layer separation and backflow, which enlarges the nozzle length to inconvenient dimensions. Furthermore, the conical converging part (either in converging–diverging or pure converging nozzles) usually produces some focal points after which particles may diverge [35, 36]. The location of these focal points depends greatly on the particle size and initial position prior to the nozzle. The sudden free expansion of the flow produces a high outward radial velocity at the beginning of the free jet close to the nozzle outlet. This radial velocity (drag) greatly varies with the radial position, both in the jet and at the nozzle outlet. It is weak at the centerline, and very strong close to the nozzle wall at the outlet. Hence, in contrast to the particles located in the central regions, those far from the axis are exposed to a strong radial drag. Consequently, the trajectories of particles concentrated on the centerline can remain approximately unchanged, while particles far from the axis may strongly diverge in the free expansion. It also should be noted that the response of the particles to these radial drags is a function of the particle size. The above-mentioned radial drag can effectively separate the particles of different sizes, especially if they are not concentrated on the centerline [37]. Small particles can follow the expanding carrier gas, while large particles persist in their original trajectories. If the particles can be concentrated on the nozzle centerline, only the Brownian diffusion (and lift-force effects

3 in case of nonspherical particles [38, 39] can perturb/spread the configuration obtained by aerodynamic focusing. Focusing the clusters on the beam center will directly improve the beam intensity, while also having an influence on the other characteristics of the supersonic part of the expansion [40]. Figure 2 shows the principle of operation of the typical apparatus used for a cluster beam, and based on a supersonic cluster source. It consists of three differentially evacuated chambers, and it can operate in the high- and ultrahigh-vacuum regimes. The first chamber hosts the cluster source; the supersonic cluster beam enters the second chamber through an electroformed skimmer. The second chamber can be equipped with a sample holder, which can intersect the beam, a quartz microbalance for beam intensity monitoring; it can also host a beam chopper or a fast ionization gauge for time-of-flight measurements of the velocity distribution of particles in the beam [32]. The third chamber usually hosts a linear time-of-flight mass spectrometer (TOF/MS), which is placed collinearly or orthogonally to the beam axis [4].

3. CARBON CLUSTER BEAM SOURCES 3.1. Laser Vaporization Cluster Source Carbon clusters can be condensed from species obtained by laser ablation of a solid target [9]. A laser vaporization cluster source (LVCS) was originally developed for the production of clusters in molecular beams by Smalley [41–43]. Since this pioneering work, laser vaporization has become one of the most common techniques for generating cluster beams of refractory materials [4]. The design and realization of LVCS are very simple (Fig. 3), especially if compared to classical hot oven sources. LVCSs are conceptually similar to the chambers used for laser ablation for thin-film deposition [44]; the only relevant difference is that the plasma plume expands in a buffer gas. In an LVCS, the light of a high-intensity pulsed laser is focused onto a target rod, and a small amount of material is vaporized into a flow of an inert carrier gas. The inert gas quenches the plasma, and cluster condensation is promoted. The mixture is then expanded in vacuum, and forms a supersonic beam. LVCSs are operated in a pulsed regime; hence, pulsed valves are used for delivering the carrier gas. This allows the use of a relatively economical apparatus with moderate pumping speed. substrate channeltron quartz microbalance

carrier gas

source cluster beam

2000 l/s diff. pump

time of flight mass spectro meter

1000 l/s diff. pump 500 l/s turbo pump

Figure 2. Schematic representation of the supersonic beam apparatus for the deposition of cluster beams. Reprinted with permission from [5], P. Milani et al., in “Nanostructured Carbon for Advanced Applications.” Kluwer, Dordrecht, 2001. © 2001, Kluwer Academic.

Cluster-Assembled Nanostructured Carbon

4 VAPORIZATION LASER

TARGET ROD Figure 3. Schematic representation of the laser vaporization cluster source (LVCS). Reprinted with permission from [45], P. Milani and W. A. de Heer, Rev. Sci. Instrum. 61, 1835 (1990). © 1990, American Institute of Physics.

Several factors affect the LVCS performance in terms of intensity, stability, and the cluster mass range attainable: quantity of ablated material, plasma–buffer gas interaction, plasma–source wall interaction, and cluster residence time prior to expansion [45]. The characteristics of the cluster population are controlled by the local gas pressure during plasma production, and by the residence time of the particles in the source body. The plasma–gas interaction affects not only the final cluster distribution, but also the subsequent expansion and beam formation. Increasing the pressure during the ablation results in a shift of the cluster distribution toward larger masses. Laser pulse characteristics influence the quantity and the state of the matter removed from the target. Pellarin et al. showed that the use of a Ti:sapphire laser (790 nm, 30 ns pulse width) can produce very intense cluster beams over a wider mass range [46]. Although the microscopic mechanisms underlying this improvement have not been characterized in detail, the longer pulse duration should be responsible for a more efficient material removal from the target [44]. Smalley and Kroto and co-workers used a disk LVCS for producing carbon clusters and C60 [47]. A modified version of this source, with a computer-controlled disk motion and a fast pulsed valve, has been coupled with a Fourier transform ion cyclotron resonance apparatus to characterize large carbon clusters in the beam [10]. Cluster-assembled carbon films have been produced using clusters generated by an LVCS by the group of Melinon [48].

3.2. Pulsed Microplasma Cluster Source The pulsed microplasma cluster source (PMCS) is based on a microplasma ablation. By confining a plasma in a welldefined region of the target, it is possible to achieve material vaporization with high efficiency. Source geometry and dimensions can be optimized to produce intense and stable cluster beams, while using a relatively low carrier–gas load [30]. Due to its extremely high intensity and stability, the PMCS is a unique tool for the synthesis of nanostructured carbon-based materials, providing long-time stability and high deposition rates [30].

The PMCS is schematically shown in Figure 4. It consists of a ceramic cubic body in which a few cubic centimeters cylindrical cavity is drilled. A channel of 6 mm diameter is drilled perpendicular to the axis of the cavity, and hosts two rods (electrodes) of the material to be vaporized, facing each other and separated by a small gap. A solenoid pulsed valve injects helium at the pressure of several tens of bars through a ceramic nozzle on the back of the cavity. The gap between the electrodes is off axis, in such a way that the gas flushes on the cathode surface. A removable nozzle with a cylindrical shape closes the front of the source. The principle of operation of the cluster source is the following: the valve delivers an intense He pulse with an opening time of a few hundreds of microseconds, thus forming a small high-gas density region at the cathode surface. After that, a very intense discharge (hundreds of amperes) lasting for few tens of microseconds is fired between the electrodes by applying a voltage ranging from 500 up to 1500 V. Ionized helium sputters a small area of cathode surface at the point where the He flux impinges on the electrode. The mixture of helium-vaporized atoms quenches, and cluster nucleation takes place. The clusters are then carried out through the nozzle by supersonic expansion [30]. Due to the source internal cavity configuration, the ablation process can be localized on a small region of the target, and hence it is very efficient. As in laser vaporization sources [45], the cathode is continuously rotated by means of an external motor in order to allow constant ablation conditions for all pulses and a homogeneous consumption of the rod. Deposition rates, up to 65 m/h over an area of 1 cm2 , can be obtained substituting the simple cylindrical nozzle with a more complex one, called a focuser [49]. Exploiting inertial aerodynamic effects [40, 49], the focuser reduces the angular semiaperture of the beam from ∼12 to less than 1 , concentrating the cluster on the center of the beam. Using the typical mass distributions of a PMCS, it is possible to grow films with a density ranging from 0.9 up to 1.4 g/cm3 , depending on the precursor clusters. BET analysis shows a surface specific area of roughly 750 m2 /g [50–52]. Time-of-flight mass spectrometry performed on carbon cluster beams shows that the length of a single pulse is about 10 ms, and the beam intensity decreases exponentially, resembling the gas pressure fall in the source. The velocity anode rod thermalization cavity pulsed valve beam

ceramic body

nozzle cathode rod

Figure 4. Expanded view of the PMCS. Reprinted with permission from [56], E. Barborini et al., Chem. Phys. Lett. 300, 633 (1999). © 1999, Elsevier.

Cluster-Assembled Nanostructured Carbon

5

of the carrier gas is about 2000 m/s in the head of the pulse, and slows down to 1000 m/s in the tail. A velocity slip of the clusters with respect to the carrier gas is present, but becomes of some relevance only for the last clusters exiting from the nozzle. The kinetic energy of the clusters is lower than 0.4 eV/atom [53]. The intensity and mass distribution of carbon clusters in the beam can be controlled and optimized by a precise synchronization between gas injection and discharge firing. Cluster intensity is also strongly dependent on the quantity of ablated material, which increases as function of the discharge voltage. Cluster mass distribution is lognormal, as usually observed in gas-phase-grown particles [21]. For typical operation parameters, it is peaked around 550 atoms/cluster, with a mean size of 850 atoms/cluster, as shown in Figure 5. The mass distribution moves toward larger masses as the residence time increases; for this reason, the population found in the front of the pulse differs from the one in the tail. Figure 5 is representative of the average spectrum obtained by integrating over the whole cluster pulse. Depending on the application, the PMCS can operate under different conditions. An ultrahigh-vacuum (UHV) PMCS has been developed in order to allow in-situ characterization of cluster-assembled carbon [54].

4. CONTROL OF FILM NANOSTRUCTURE BY CLUSTER SELECTION The assembling of carbon cluster beams has been investigated both experimentally and theoretically in view of the creation of a new class of solids based on nanoscopic structures (chains, fullerene-like units, etc.), otherwise difficult to produce by assembling the film atom by atom [6, 18, 31, 55, 56]. This could be achieved by control of the cluster size distribution prior to deposition. Novel structural and functional properties could be displayed by solids, retaining some of the cluster characteristics [6, 18]. The growth of carbon films via low-energy cluster deposition can be viewed as a random stacking of particles as for

4

ballistic deposition [53, 57]. The resulting material is characterized by a low density compared to that of films assembled atom by atom, and it shows different degrees of order, depending on the scale of observation [25, 26, 31]. The morphology of cluster-assembled carbon films shows different degrees of order and structures, depending on the scale of observation. The characteristic length scales are determined by cluster dimensions and by their fate after deposition: cluster beams are characterized by the presence of a finite mass distribution and by the presence of different isomers with different stabilities and reactivities. Once on the substrate, stable clusters can survive almost intact, while reactive isomers can coalesce to form a more disordered phase. Classical molecular dynamics simulations tailored for SCBD experiments have been used to characterize and predict the growth of carbon thin films assembled by SCBD [18]. The simulations show that, depending on the cluster mass distribution, films with fullerene structures embedded in an amorphous matrix characterized by threefold coordination and hexagonal bond rings are obtained with large clusters (Fig. 6). Smaller and more fragile clusters such as linear chains produce disordered structures characterized by nonhexagonal rings with twofold coordinated atoms. Structural and elastic properties are in good agreement with the experiments. The large porosity allows the accommodation of mechanical stresses, thus permitting the growth of films with very large thicknesses (>10 m) [29, 53]. Molecular dynamics simulations have also been performed to describe the deposition of small fullerene-like carbon clusters. C28 [19] and C20 [58] have been used as elemental building blocks. The growth of hydrocarbon thin films has also been simulated, starting from the deposition of small clusters at various kinetic energies [59].

× 109

3.5

Intensity (arb.u.)

3 2.5 2 1.5 1 0.5

(a)

0 –0.5 0

500

1000

1500

2000

2500

3000

3500

atoms/cluster Figure 5. Typical mass spectrum for a supersonic beam of carbon cluster produced by a PMCS. Mass distribution is peaked at about 550 carbon atoms/cluster, with a mean size of 850 atoms/cluster.

(b)

Figure 6. Atomic structure of amorphous film obtained by deposition of beams with bimodal cluster mass distribution in the range 1–23 atoms/cluster and 46–120 atoms/cluster with relative weight 5:1 [sample (a)] and 1:10 [sample (b)], respectively. Reprinted with permission from [18], D. Donadio et al., Phys. Rev. Lett. 83, 776 (1999). © 1999, American Physical Society.

6

5. STRUCTURAL AND FUNCTIONAL CHARACTERIZATION 5.1. Raman Scattering An enormous number of papers have been devoted to the Raman characterization of different forms of disordered and nanocrystalline carbon, and many reviews can be found on this subject [60–63]. Raman scattering is perhaps the most popular method for the routine characterization of carbon-based materials [64, 65]. From Raman spectra, it is, in principle, possible to detect the coexistence of different carbon allotropic forms, the degree of order of different structures, the nature of the local bonding, and to follow the structural evolution as a function of the physicochemical treatments of the samples. The popularity of Raman scattering for the study of carbon, however, does not correspond to well-established models for the interpretation of all of the features observed in carbon-based materials: up to now, especially for disordered carbonaceous materials, no satisfactory theory of the Raman effect has been provided [60]. Raman scattering is very sensitive to structural changes that modify the translational symmetry of the perfect crystal. Different carbonaceous materials, characterized by the degree of disorder, ranging from the atomic scale, typical of noncrystalline systems, to nano- and microcrystalline assemblies, should display different Raman spectra. The coexistence of the contributions coming from localization effects and structural disorder renders it very difficult to extract, from Raman spectra of disordered carbon, information on the structure and dimensions of domains. In particular, the sp2 /sp3 ratio in amorphous and diamondlike carbon cannot be derived solely from Raman spectra analysis [60]. Raman scattering can be easily interpreted in terms of crystallites dimensions when G and D bands are well separated and distinct. These bands have relatively narrow bandwidths, whereas “nonorganized” or disordered carbon contributes to their broadening. In the case of large broadening and merging of the G and D bands, the extraction of structural information is very difficult, and interpretation of the Raman spectra must be performed very carefully to disentangle the different contributions to the line shapes. Several factors, such as crystallite dimensions and size distribution, bond angle disorder, and stress, affect the Raman line shape with a weight, which is predictable with difficulty [65]. In amorphous carbon, the G and D bands broaden and coalesce; the Raman spectra can be interpreted as being originated by an amorphous carbon matrix of threefold and fourfold coordinated carbon atoms in which a small region of hexagonal rings is embedded [65]. Different synthetic methods can change the sp2 /sp3 ratio [63]; however, Raman spectra only exhibit small changes, confirming that they are mainly sensitive to the sp2 content. The Raman characterization of cluster-assembled materials must address the problem of cluster coalescence and their organization in structures spanning length scales from the nanometer up to the micrometer. The different structures in which the precursor clusters are organized need

Cluster-Assembled Nanostructured Carbon

experimental probes sensitive to the different length scales typical of intracluster and intercluster interactions. For carbon-based materials, Raman spectroscopy can be used for the characterization on a nanometer scale [66]. Melinon and co-workers [33, 48] characterized the structural and mechanical properties of cluster-assembled carbon films obtained with an LVCS at different operating conditions, which should correspond to different cluster mass distributions. The observation of mass spectra peaked at C20 , C60 , and C900 has been interpreted as an indication of different size distributions; however, a systematic characterization of the cluster beams has not been made by taking into account the residence time of the clusters in the source. Low-density films (1 g/cm3 , characterized by a granular structure on a scale of several tens of nanometers, show Raman spectra evolving from a graphitic character when large clusters (C900 are used to an amorphous character as the cluster size distribution is reduced to a few tens of atoms per cluster [48]. Melinon and co-workers assumed, from Raman spectra taken with excitation at 532 nm, that films formed with small clusters have an sp3 character due to the formation of a diamond-like phase by the assembling of cage-like C20 clusters. Unfortunately, the available experimental evidence does not support this hypothesis. Despite the uncertainty on the final film structures, these studies showed the existence of a “memory effect” that is the dependence of the final structure of the film on the mass distribution of the precursor clusters [48]. Raman characterization has been performed on nanostructured carbon films deposited with clusters of several hundreds atoms at low kinetic energy by a PMCS [31, 55, 56, 66]. The original graphitic structure of the clusters was essentially conserved after the landing on the surface, producing a porous graphite-like film. Micro-Raman spectroscopy indicates that the films are formed at a submicrometric scale by two coexisting phases, one amorphous and the other one more graphitic. Figure 7 shows a typical micro-Raman spectrum of a clusterassembled film. The spectrum can be described as a superposition of contributions coming from amorphous carbon (B) and disordered graphite (C). These two phases are uniformly distributed, and coexist over the whole sample. The more ordered phase probably originates from the aggregation of large graphitic clusters prior to deposition. Macro-Raman spectra of macroscopic portions of the films show an increase of the graphitic character when the cluster growth conditions favor the formation of large particles. In these films, the graphitic microregions are more abundant. By changing the PMCS operation parameters, different structures have been grown (Fig. 8). The two main Raman features at about 1550 and 1350 cm−1 can be, respectively, identified as the G-band peak of crystalline graphite and the D-band peak due to crystal disorder. It is evident that source parameters (in this case, the amount of He filling PMCS, increasing from A to C) change cluster size distribution and determine different film structures. Spectrum A is typical of amorphous sp2 carbon, having a broad G structure shifted at 1540 cm−1 (G peak of crystalline graphite appears at 1581 cm−1 and just a shoulder as a D structure. Spectrum C shows a well-defined D peak, and a narrow G peak

Cluster-Assembled Nanostructured Carbon

7

300

250

Intensity (a.u.)

200

150

A

100

B 50

C 0 500

1000

1500

2000

2500

3000

Raman shift (cm–1) Figure 7. (A) Micro-Raman spectra of ns–C film. This is a superposition of a spectrum typical of amorphous carbon (B), and one typical of polycrystalline graphite (C) with defined D and G bands. Spectrum (C) was obtained by subtracting (B) from (A). Reprinted with permission from [103], C. Ducati et al., J. Appl. Phys. 92, 5482 (2002). © 2002, American Institute of Physics.

centered at 1581 cm−1 . This is consistent with a disordered graphitic structure. Different film structures are related to different relative abundances of small (few atoms) clusters and larger fullerene-like clusters. Keeping constant the thermodynamic conditions in the cluster source, it has been demonstrated that a selection of clusters can be made in the beam by using aerodynamical gas effects [56]. It is possible to separate light clusters from the heavy ones by forcing gas flux lines to make sudden

Intensity (a.u.)

200

A

turns. The light particles will be able to follow the flux line trajectories, whereas the large ones will maintain their original trajectories. By varying the background pressure in the source chamber, a shock wave can be produced in front of the skimmer, causing mass separation effects and changing the final characteristics of the beam (Fig. 9). In particular, due to their different inertia, light species follow diverging streamlines after the shock front, while heavy species are not diverted and can follow straight trajectories through the skimmer. Due to this effect, large clusters are concentrated in the central portion of the beam, whereas the lighter ones are at the periphery. By placing substrates in different portions of the beam, films from different precursor clusters can be grown [56]. Figure 10 reports Raman spectra of films deposited with different portions of the beam. The top spectrum is taken on a region rich with small clusters deposited with the periphery of the beam. The shape and the shift of the peak (G band) are typical of amorphous sp2 carbon. The D peak is present as a broad shoulder on the left of the G band. Going from the top to the bottom spectrum in Figure 10, one can follow the evolution from an amorphous toward a disordered graphitic structure by shifting the mass distribution from small to large clusters (from the periphery to the center of the beam). This evolution is confirmed by hardening of the G peak, the appearance of a well-defined D band, and by the narrowing of the two lines. All of these parameters are in agreement with a graphitization of the sample [60]. Cluster separation can also be obtained by using an appropriate nozzle configuration (focusing nozzle) [49]. Films grown using a cluster mass distribution depleted by the large fullerene-like clusters are expected to show a very disordered structure; on the other hand, films grown with a cluster distribution rich in the fullerene-like component should be characterized by a disordered graphitic structure. Raman spectra show that films deposited with large clusters are characterized by the presence of a high number of distorted sp2 bonds, whereas when depositing small clusters (few tens of atoms), sp3 and sp coordination is present. These results are in qualitative agreement with XPS measurements [67], although photoemission spectroscopy is

150

detached shock

B 100

skimmer

C beam

heavy particle trajectories

50

500

1000

1500

2000

2500

3000

Raman shift (cm–1) Figure 8. Micro-Raman spectra of ns–C films showing the influence of the PMCS helium pressure during cluster formation. The He pressure increases from (A) to (C) from 8 to 16 torr, and this parameter controls the number of large graphitic nanoparticles present in the beam. Reprinted with permission from [103], C. Ducati et al., J. Appl. Phys. 92, 5482 (2002). © 2002, American Institute of Physics.

light particle trajectories Figure 9. Varying the background pressure in the source chamber, a shock wave can be produced in front of the skimmer, causing mass separation effects and changing the final characteristics of the beam. Reprinted with permission from [56], E. Barborini et al., Chem. Phys. Lett. 300, 633 (1999). © 1999, Elsevier.

Cluster-Assembled Nanostructured Carbon

8

Intensity (a.u.)

periphery

center

500

1000

1500

2000

Raman shift (cm–1) Figure 10. First-order Raman spectra of films deposited with different cluster mass distributions in the beam. Selection is obtained by using different regions of the beam. In the center, there is a concentration of large clusters, whereas small clusters are at the periphery. Going from the bottom (internal part of the beam, i.e., large clusters) to the top spectrum (external part of the beam, i.e., small clusters), one can see the evolution from a disordered graphitic structure to an amorphous one. Reprinted with permission from [56], E. Barborini et al., Chem. Phys. Lett. 300, 633 (1999). © 1999, Elsevier.

sensitive to a scale smaller than that of Raman. For large clusters, optical spectra confirm the disordered sp2 structure, and show the presence of a gap of 0.6 eV [55].

5.2. Brillouin Scattering Although Raman spectroscopy is a powerful tool for studying the structure of carbon-based materials, the information is restricted to the scale of a few nanometers. The experimental characterization and theoretical modeling of clusterassembled granular materials must address the problem of cluster coalescence and their organization in structures spanning length scales from the nanometer up to the micrometer. The different structures in which the precursor clusters are organized needs experimental probes sensitive to the different length scales typical of intracluster and intercluster interactions. In order to study the organization of clusters on a scale of hundreds of nanometers, which is the typical scale of thermally excited long-wavelength acoustic phonons, Brillouin light scattering can be used [52]. Films of graphite [68], polycrystalline diamond [69], diamond-like a–C:H [70], fullerite [71], and phototransformed C60 [72] have also been studied by Brillouin scattering. From Brillouin spectra, the elastic properties of the material can also be extracted with a higher accuracy compared to nanoindentation measurements, which require a complex analysis and a careful interpretation. Bulk and surface Brillouin scattering signals have been obtained from cluster-assembled carbon films characterized by a complex structure from the atomic to the hundreds of

nanometers level, showing that this technique also can be used for nanostructured materials with irregular surfaces. The bulk and shear modulus of the material have been determined, giving information on the acoustic properties on a mesoscopic scale. This allows us to infer the nature of the bonding between the carbon aggregates. Cluster-assembled films are characterized by a mesostructure with low density and high porosity, originated by the coalescence of the primeval clusters. Films produced with large clusters show a surface roughness higher than films produced with small clusters, with comparable thickness. Moreover, roughness evolves more rapidly with respect to thickness in films produced with large clusters. Films produced with small clusters show a more close-packed structure, as revealed by AFM measurements [73]. Thick films (thickness >05 m) and thin films (thickness 100)000). In addition to the high specific power, the energy storage in supercapacitors is simpler and more reversible than in conventional batteries [138]. Supercapacitors are formed by two polarizable electrodes: a separator and an electrolyte. The electrical field is stored at the interfaces between the electrolyte and the electrodes. The layer of the charges in the electrode and the layer of solvated ions of opposite sign constitute the so-called “double layer.” The dielectric medium is the high-permittivity electrolyte. The electrode surface is extremely huge, and the distance between the opposite charges is in the nanometer scale due to the double layer. The double-layer phenomenon at the boundary between an ionic liquid and a solid was been discovered in 1879 by Helmholtz, and studied by several authors [139]. A theoretical treatment can be easily found in electrochemistry textbooks, for example, [140, 141]. At the surface of a metal in an electrolyte, the following exchange reaction occurs, depending on the energies of the different species in the electrode and in the electrolyte: Mez+ + Ze− ↔ Me Here, MeZ+ is the solvated metal ion (Z indicates the valence), Ze− are the exchanged electrons, and Me is the solid metal constituting the electrode. The positive or negative charges in the electrode are compensated by a layer of cations or anions adsorbed at the surface of the electrode (the double layer). If a potential is applied, the charge of the double layer can be changed, that is, increased or decreased, or even the sign of the charge may change. If the electrode is immersed in an inert electrolyte, no reaction can occur upon a potential change; only the double layer is charged. There is no charge transfer between the electrode and an electrolyte as long as the potential is smaller than the dissociation potential of the electrolyte. As reported in Figure 29, the double-layer thickness is defined as half the diameter (a/2 of the adsorbed solvated ions. This simple model of the Helmholtz double layer is called a rigid double layer [142], and corresponds to a twoplate capacitor where the plates are at a distance of a/2.

Cluster-Assembled Nanostructured Carbon

21

rigid double-layer

diffuse double-layer

a/2 Potential ϕMe ∆ϕrigid

ϕo.H.

∆ϕdiff. ϕEl. x=0 ξ=0

ξ=χ

ξ, x

Figure 29. Helmholtz double layer in the approximation of rigid layer. Correspondingly, the electric potential is reported as a function of the distance from the electrode surface: +Me is the electrode potential, +oH is the potential at the outer Helmholtz layer, and +El is the potential of the electrolyte. Reprinted with permission from [142], O. Stern, Z. Electrokem. 30, 508 (1924). © 1924,

The potential decreases linearly with distance to the electrolyte potential within the adsorbate rigid layer (also called the Stern layer). Due to thermal diffusion, the potential decreases exponentially outside the outer Helmholtz surface (, = 0). At , = -, the potential is decreased to 1/e of the total potential (.+ = .+rigid + .+diff . The value of - is approximately 10 nm, and decreases with increasing ionic concentration of the electrolyte. For an electrolyte concentration greater than 0.1 mol/l, - decreases to the outer Helmholtz surface (0.1 nm). Therefore, the Helmholtz double layer is a capacitor in which the charges of different sign are separated by just a few nanometers: this allows EDLC to have a high capacity. The working voltage of an ECDL capacitor is determined by the decomposition voltage of the electrolyte. For a single cell and an organic electrolyte, it is approximately 2.5 VDC . In addition, the operating voltage depends mainly on the environmental temperature, the current intensity, and the required lifetime. The capacitance of an ECDL capacitor can be very large, for example, several thousands of farads, thanks to the very small distance which separates the opposite charges at the interfaces between the electrolyte and the electrodes, and thanks to the very huge surface of the electrodes. These capacitances are several orders of magnitude greater than that of conventional capacitors. The goal of ECDL capacitor research is to reach the highest energy and power densities to get the smallest component volume and weight for a given application. These goals have been fixed by the U.S. Department of Energy (DOE) to 15 W·h/kg for the energy density and 2000 W/kg for the power density. To meet these values, it is necessary to increase the capacitance density, to reduce the series resistance (see below), and to increase the cell operating

voltage. The efforts are concentrated on the study of the electrode surface, the electrode accessibility, and the electrolyte decomposition voltage. Nanostructured materials based on the assembly of carbon nanotubes are attracting increasing interest in view of energy storage applications [143]. Several requirements still remain to be fulfilled before the development of a nanotube-based technology, in particular, the production and manipulation of macroscopic quantities of nanotubes with controlled structure and dimensions. In this sense, if compared with other applications, electrochemical ones are less demanding since a controlled nanostructure is required without internal atomic perfection. The use of nanotube mats and, in general, of nanostructured carbon materials thus offers great potential for electrochemistry and, in particular, for the production of porous electrodes for supercapacitors. Since the capacitance is proportional to the surface area, electrochemical inert materials with the highest specific surface area are utilized for supercap electrodes in order to form a double layer with a maximum number of electrolyte ions [144]. Therefore, carbonaceous materials in their various forms have been studied as electrodes for the construction of rechargeable ECDL capacitor energy storage devices. Activated carbon composites and fibers [145] with surface areas up to 2000 m2 /g (measured by gas absorption), as well as carbon aerogels [138] with surface areas up to 850 m2 /g have been investigated. Unfortunately, a significant part of the surface area resides in nanopores (sizes below 2 nm) which are not accessible to the electrolyte ions. The surface area that remains electrochemically accessible (mesopores, macropores) reveals capacitance values well below the estimated values. Therefore, the pore size distribution, together with the surface area, are important for the determination of the double-layer capacitance. Niu et al. [143] realized supercap electrodes with freestanding mats of catalitically grown nanotubes. The nanotube electrodes (diameter of 0.5 in) have shown a surface area of 430 m2 /g, and a measured specific capacitance up to 113 F/g with an aqueous electrolyte, resulting in a power density above 8 kW/kg. Although very promising, the use of nanotubes in practical devices is limited by the complex production/purification method, the small area of the freestanding nanotube films, and the adhesion of nanotube mats to large surfaces which have to be wound into a capacitor. The morphological features of nanostructured carbon films match the requirements on supercap electrodes. The hierarchical organization of clusters in grain structures covering a wide range of length scales (from nanoscopic up to mesoscopic) accounts for high porosity and low density, and particularly fulfills the demand for a high specific surface. As other forms of carbon, this material appears to be inert with respect to the electrochemical processes involved in a supercap. Difficulties could arise from the high resistivity of cluster-assembled nanostructured carbon film [146], which limits the power density. With respect to nanotube-based technology, requiring a complex sequence of steps to obtain the electrodes from the as-grown material, supersonic cluster beam deposition appears to be a simpler and versatile technique. Moreover, the presence in the beam of clusters of different masses

Cluster-Assembled Nanostructured Carbon

22 could allow the selection of precursors, thus modifying the properties of the deposited electrode. Granularity and conductivity should be the main properties to be controlled. A single ECDL cell device was fabricated with two nanostructured carbon electrodes with a thickness d = 08 m and a surface of 6 cm × 6.6 cm (for each electrode) deposited on an aluminum substrate (30 m thick) serving as a collector. The two carbon electrodes were separated by a 36 m thick polymer separator as a membrane. The cell is impregnated using a quartenary ammonium salt soluted in propylene carbonate (PC) as an organic electrolyte. Impregnation was performed under controlled conditions (argon inert atmosphere) in a glove box. The carbon-coated electrodes and the separator were stacked between two 3 mm thick aluminum plates. Finally, the stack was pressed together for better contact, and was sealed. A dc voltage is used to charge the supercapacitor to its nominal voltage V0 = 27 V. The capacitance C and the equivalent series resistance ESR are determined by measuring the voltage over the contacts of the ECDL cell during discharge over an external circuit resistance Re of 100 . A schematic view of the cell and the setup for the C and the ESR measurement methodology is shown in Figure 30. The equivalent circuit of an ECDL capacitor consists of a capacitance C, a parallel resistance Rp responsible for the self-discharge, an inductivity L, and the equivalent series resistance ESR representing the internal resistance of the capacitor. The series resistance keeps the current low, and therefore limits the power of the component and is responsible for the electric losses. The key factors determining the high energy and power density are the capacity (maximum C is required) and the series resistance (minimal ESR is required), respectively. The time constants of the charging and discharging circuits are equal to 0Re + ESR × C, while the time constant of the self-discharge is equal to Rp × C. In order to minimize the self-discharge, Rp should be as large as possible. The electrolyte itself strongly influences the capacity as well as the series resistance of the ECDL capacitor. Tanahashi et al. [145] did a comparison between aqueous and organic electrolytes, showing on one side that organic electrolytes can operate at higher voltages (up to 3 V compared to 1.2 V for aqueous electrolyte), desirable for high energy and power densities. On the other side, the aqueous electrolyte reveals a higher capacity and an ESR lower by a factor of 10. The lower ESR of the aqueous electrolyte, desirable for high-power densities, is due to the higher conductance of the ions, resulting in a voltage drop by a factor Re

I0

separator electrolyte electrode Al foil Al support

Rs Ut C

Rp

Figure 30. Schematic view of an ECDL cell and the setup for the capacity C and the equivalent series resistance ESR measurement methodology.

of 10 lower than for organic electrolytes at the beginning of the capacitor discharge. During tests, the nanostructured carbon ECDL prototype was discharged over an external circuit resistance Re . The voltage drop .U at the beginning of the discharge of the capacitor allows us to determine ESR, while C is determined on the basis of ESR after 20 and 30 s of discharge. The capacity C and the equivalent series resistance ESR were determined with the following equations: C=

ESR =

−t 

0ESR + Re · ln 0ESR + Re · Re .U = .U · it=0 U1) t=0 − .U

Ut U1) t=0 ·Re



where t is the time, and Ut is the voltage at time t. For the single-cell device, the best results are ESR = 113 ± 009  and C = 0099 ± 0004 F; per surface unit: ESR = 457 /cm2 and C = 0003 F/cm2 . Based on these values and with a density of 1 g/cm3 , the specific capacity of one electrode is c = 2C/d = 75 F/g. The maximal energy 2 /04 × ESR , and with its nomidensity is given by P = Umax nal voltage V0 = 27 V, it results in 76 W·h/kg. The maximal power density is 506 kW/kg [147]. These results demonstrate that cluster-assembled carbon films are very attractive for supercapacitor realization. The optimization of parameters such as electrolyte accessibility, film conductivity, and interface resistance is currently underway. The scale up of the cluster deposition parameters to match industrial requirements will also be explored.

6.6. Nano- and Microfabrication with Carbon Cluster Beams The fabrication of functional devices based on nanostructured materials requires the ability to assemble nanoparticles in micrometer and submicrometer patterns with high precision and compatibility with planar technology [148]. Physical and chemical deposition techniques are used for the patterning of semiconductor, metallic, and polymeric films and the production of dot arrays. These techniques require putting resists or stamps in contact with a substrate that undergoes different pre- or postdeposition etching or thermal treatments [149–151]. The effect of etching or thermal treatments on clusterassembled materials has not been investigated systematically; however, several problems due to the “granularity” of nanostructured materials can be expected. For nanostructured films, a noncontact patterning has been demonstrated based on the principle of depositing particle beams on a substrate through a stencil mask [152]. Due to the use of supersonic cluster beams, this process has a high lateral resolution, large area deposition, high deposition rate, and low temperature processing [153]. SCBD can be used for the fabrication of patterns of micrometer-size cluster-assembled objects. By exploiting the high intensity and collimation typical of supersonic expansions, it is possible to create structures with a high aspect ratio and controlled shape, arranged in ordered arrays on any kind of substrate at room temperature.

Cluster-Assembled Nanostructured Carbon (A)

(B)

Figure 31. (A) SEM images of ns–carbon pillars deposited on a polymeric substrate. A standard metallic TEM grid with round holes of 20 m diameter is used as a mask. The mask–substrate distance is about 0.3 mm. (B) The same sample is tilted to show the 3-D nature of the deposit.

Figure 31 shows SEM images of a pattern of nanostructured carbon pillars obtained by deposition of carbon clusters. The pattern has been obtained by using a round hole grid. Atomic force microscope analysis reveals granularity at a nanometer scale based on the assembling of grains with a mean diameter of 20 nm. Dots with complex geometries and a high aspect ratio, as well as self-standing microstructures can be obtained with this technique. Key issues for micropatterning applications of SCBD are process speed, resolution, surface finish, and substrate compatibility [152]. Supersonic cluster beam sources are characterized by high directionality and high intensity. These properties have been used for the thin-film metallization of flat and microstructured surfaces [152]. SCBD allows the creation of micrometer-size objects having a structure on the nanoscale. The stencil mask is not in contact with the substrate; patterns on different substrates such as silicon, aluminum, copper, stainless steel, and polyethylene can be deposited at room temperature. Lateral resolution depends not only on the sharpness of the mask edges, but also on the beam collimation. Supersonic expansions have a higher collimation compared to effusive beams [49]; however, clusters in supersonic beams can still have a nonnegligible thermal divergence and a broad mass distribution [49]. With a standard supersonic beam, a step produced with a straight edge mask is very noisy, and not well defined [49]. In order to increase the step sharpness, one should increase the degree of supersonicity of the beam, but this will require a huge pumping system [4, 49]. The use of a focusing nozzle can circumvent this problem by concentrating the clusters on the center beam, keeping intact the other expansion parameters. A step with a width of 450 nm using a mask placed at 0.33 mm from the substrate has been reported.

7. APPLICATIONS OF CLUSTER-ASSEMBLED CARBON Nanostructured carbon-based materials and, in particular, those assembled from nanotubes, are attracting an increasing interest in view of a wide variety of applications ranging from micromechanics to electronics and energy storage [154]. The use of cluster-assembled carbon films can be an interesting alternative for applications in the abovementioned fields. Nanostructured carbon films are good field emitters, showing reasonably high current densities and stability [101, 103]. The use of supersonic beams allows deposition at

23 very high rates over sizable areas at room temperature. This is a considerable advantage since it allows us to overcome limitations for the thermal sensitivity of many substrates. Compared to other emissive carbon films, clusterassembled carbon films can be deposited, for example, onto glass substrates. In contrast, nanotubes can so far be formed only under high-temperature conditions [155]. Pulsed-laser-induced electron emission has also been observed from ns–carbon films. These properties may be of interest for the construction of photoemitting cathodes with applications for accelerator injectors [156]. Due to its high porosity and highly accessible surface area, ns–carbon is a good candidate for the production of supercapacitors [147]. Films with a density of 1 g/cm3 have shown, in a dc regime, a specific capacitance per electrode of 75 F/g and energy and power densities of 76 W/kg and 506 kW/kg, respectively [147]. ns–carbon films, once deposited, do not require further chemical purification or thermal annealing. The use of cluster beams allows the deposition of films on thin foil films compatible with the actual manufacturing processes for supercapacitors [157]. Cluster-assembled carbon grown by pulsed laser deposition has been used for the production of preconcentrators for gas trace analysis by gas chromatography microsensors [29].

GLOSSARY Cluster Aggregate of atoms, whose number of constituents can vary from a few units to many thousands. Nanostructured materials Materials whose building blocks, instead of single atoms, are nanometer-sized units (i.e. clusters). (Supersonic) molecular beam A stream of molecules or atoms, possessing high directionality, propagating in a vacuum environment. In the case of supersonic beams, mean velocity is larger than the local velocity of sound. Supersonic cluster beam deposition (SCBD) Technique for the deposition of films exploiting a supersonic beam of clusters. The film is grown in high or ultra-high vacuum, intercepting the cluster beam with a suitable substrate.

REFERENCES 1. A. M. Rao and M. S. Dresselhaus in “Nanostructured Carbon for Advanced Applications” (G. Benedek, P. Milani, and V. G. Ralchenko, Eds.), NATO Science Series, Vol. 24. Kluwer, Dordrecht, 2001. 2. H. S. Nalwa, Ed., “Handbook of Nanostructured Materials and Nanotechnology.” Academic Press, San Diego, 2000. 3. R. E. Clausing, L. L. Norton, J. C. Angus, and P. Koidl, Eds., “Diamond and Diamond-like Films and Coatings,” NATO ASI Series B: Physics 266. Plenum, New York, 1991. 4. P. Milani and S. Iannotta, “Cluster Beam Synthesis of Nanostructured Materials.” Springer, Berlin, 1999. 5. G. Benedek, P. Milani, and V. G. Ralchenko, Eds., “Nanostructured Carbon for Advanced Applications,” NATO Science Series, Vol. 24. Kluwer, Dordrecht, 2001. 6. L. Ravagnan, F. Siviero, C. Lenardi, P. Piseri, E. Barborini, P. Milani, C. Casari, A. Li Bassi, and C. E. Bottani, Phys. Rev. Lett. 89, 28 (2002).

Cluster-Assembled Nanostructured Carbon

24 7. J. B. Donnet, R. C. Bansal, M. J. Wang, Eds., “Carbon Black.” Marcel Dekker, New York, 1993. 8. E. A. Rohlfing, D. M. Cox, and A. Kaldor, J. Chem. Phys. 81, 3322 (1984). 9. R. F. Curl and R. E. Smalley, Science 242, 1017 (1988). 10. S. Maruyama, R. L. Anderson, and R. E. Smalley, Rev. Sci. Instrum. 61, 3686 (1990). 11. E. A. Rohlfing, J. Chem. Phys. 89, 6103 (1988); J. Chem. Phys. 93, 7851 (1990). 12. W. R. Creasy and J. T. Brenna, J. Chem. Phys. 92, 2269 (1990). 13. G. von Helden, M. T. Hsu, P. R. Kemper, and M. T. Bowers, J. Chem. Phys. 95, 3835 (1991). 14. G. von Helden, P. R. Kemper, N. G. Gotts, and M. T. Bowers, Science 259, 1300 (1993). 15. J. M. Hunter, J. L. Fye, and M. F. Jarrold, J. Chem. Phys. 99, 1785 (1993). 16. J. M. Hunter, J. L. Fye, and M. F. Jarrold, Science 260, 784 (1993). 17. J. M. Hunter, J. L. Fye, E. J. Roskamp, and M. F. Jarrold, J. Phys. Chem. 98, 1810 (1994). 18. D. Donadio, L. Colombo, P. Milani, and G. Benedek, Phys. Rev. Lett. 83, 776 (1999). 19. A. Canning, G. Galli, and J. Kim, Phys. Rev. Lett. 78, 4442 (1997). 20. H. Haberland, Ed., “Clusters of Atoms and Molecules.” Springer, Berlin, 1994. 21. J. M. Bonard and A. Chatelain, Eds., “Small Particles and Inorganic Clusters.” Springer, Berlin, 1999. 22. G. A. J. Amaratunga, M. Chowalla, and C. J. Kiely, I. Alexandrou, R. Aharonov, and R. M. Devenish, Nature 383, 321 (1996). 23. T. Hihara and K. Sumiyama, J. Vac. Sci. Technol. B 17, 1923 (1999). 24. I. Alexandrou, H. J. Scheibe, C. J. Kiely, A. J. Papworth, G. A. J. Amaratunga, and B. Schultrich, Phys. Rev. B 60, 10903 (1999). 25. A. V. Rode, S. T. Hyde, E. G. Gamaly, R. G. Elliman, D. R. McKenzie, and S. Bulcock, Appl. Phys. A 69, S755 (1999). 26. A. V. Rode, E. G. Gamaly, and B. Luther-Davies, Appl. Phys. A 70, 135 (2000). 27. A. V. Rode, B. Luther-Davies, and E. G. Gamaly, J. Appl. Phys. 85, 4222 (1999). 28. G. A. M. Reynolds, Z. H. Wang, M. S. Dresselhaus, A. W. P. Fung, and R. W. Pekala, Phys. Rev. B 49, 15027 (1994). 29. M. P. Siegal, D. L. Overmeyer, R. J. Kottenstette, D. R. Tallant, and W. G. Yelton, Appl. Phys. Lett. 80, 3940 (2002). 30. E. Barborini, P. Piseri, and P. Milani, J. Phys. D: Appl. Phys. 52, L105 (1999). 31. P. Milani, P. Piseri, E. Barborini, A. Podestà, and C. Lenardi, J. Vac. Sci. Technol. A 19, 2025 (2001). 32. G. Scoles, Ed., “Atomic and Molecular Beam Methods.” Oxford University Press, Oxford, 1988. 33. A. Perez, P. Melinon, V. Dupuis, P. Jensen, B. Prevel, J. Tuaillon, L. Bardotti, C. Martet, M. Treilleux, M. Broyer, M. Pellarin, J. L. Vaille, B. Palpant, and J. Lerme, J. Phys. D: Appl. Phys. 30, 709 (1997). 34. F. Huisken, B. Kohn, and V. Paillard, Appl. Phys. Lett. 74, 3776 (1999). 35. J. Fernandez de la Mora and P. Riesco-Chueca, J. Fluid. Mech. 195, 1 (1988). 36. J. Fernandez de la Mora, N. Rao, and P. H. McMurry, J. Aerosol Sci. 21, 889 (1990). 37. V. H. Reis and J. B. Fenn, J. Chem. Phys. 39, 3240 (1963). 38. P. Liu, P. J. Ziemann, D. B. Kittelson, and P. H. McMurry, Aerosol Sci. Technol. 22, 293 (1995). 39. P. Liu, P. J. Ziemann, D. B. Kittelson, and P. H. McMurry, Aerosol Sci. Technol. 22, 314 (1995). 40. H. Vahedi Tafreshi, P. Piseri, S. Vinati, E. Barborini, G. Benedek, and P. Milani, Aerosol Sci. Technol. 36, 593 (2002). 41. T. G. Dietz, M. A. Duncan D. E. Powers, and R. E. Smalley, J. Chem. Phys. 74, 6511 (1981).

42. 43. 44. 45. 46.

47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62.

63. 64. 65. 66. 67. 68. 69. 70. 71.

72. 73. 74.

75.

76. 77.

R. E. Smalley, Laser Chem. 2, 167 (1983). R. E. Smalley, Acc. Chem. Res. 25, 98 (1992). J. C. Miller, Ed., “Laser Ablation.” Springer, Berlin, 1994. P. Milani and W. A. de Heer, Rev. Sci. Instrum. 61, 1835 (1990). M. Pellarin, E. Cottancin, J. Lermé, J. L. Vialle, J. P. Wolf, M. Broyer, V. Paillard, V. Dupuis, A. Perez, J. P. Perez, J. Tuaillon, and P. Melinon, Chem. Phys. Lett. 224, 338 (1994). H. V. Kroto, J. R. Heath, S. C. O’Brien, R. F. Curl, and R. E. Smalley, Nature 318, 162 (1985). V. Paillard, P. Melinon, V. Dupuis, J. P. Perez, A. Perez, and B. Champagnon, Phys. Rev. Lett. 71, 4170 (1993). P. Piseri, A. Podestà, E. Barborini, and P. Milani, Rev, Sci. Instrum. 72, 2261 (2001). C. Lenardi, E. Barborini, V. Briois, L. Lucarelli, P. Piseri, and P. Milani, Diamond Rel. Mater. 10, 1195 (2001). C. E. Bottani, A. C. Ferrari, A. Li Bassi, P. Milani, and P. Piseri, Europhys. Lett. 42, 431 (1998). C. S. Casari, A. Li Bassi, C. E. Bottani, E. Barborini, P. Piseri, A. Podestà, and P. Milani, Phys. Rev. B 64, 085417/1 (2001). P. Milani, E. Barborini, P. Piseri, C. E. Bottani, A. Ferrari, and A. Li Bassi, Eur. Phys. J. D9, 63 (1999). E. Barborini, F. Siviero, S. Vinati, C. Lenardi, P. Piseri, and P. Milani, Rev. Sci. Instrum. 73, 2060 (2002). P. Milani, M. Ferretti, P. Piseri, C. E. Bottani, A. Ferrari, A. Li Bassi, G. Guizzetti, and M. Patrini, J. Appl. Phys. 82, 5793 (1997). E. Barborini, P. Piseri, A. Li Bassi, A. Ferrari, C. E. Bottani, and P. Milani, Chem. Phys. Lett. 300, 633 (1999). A. L. Barabasi and H. E. Stanley, “Fractal Concepts in Surface Growth.” Cambridge University Press, Cambridge, 1995. A. J. Du et al., Phys. Rev. B 66, 035405 (2002). T. A. Plaisted, B. Ni, J. D. Zahrt, and S. B. Sinnott, Thin Solid Films 381, 73 (2001). J. Robertson, Prog. Solid State Chem. 21, 199 (1991). J. Robertson, Semicond. Sci. Technol. 18, S12 (2003). M. W. Geis and M. A. Tamor, “Diamond and Diamondlike Carbon,” Encyclopedia of Applied Physics, Vol. 5, VCH Publishers, Weinheim, 1993. D. R. McKenzie, Rep. Prog. Phys. 59, 1611 (1996). M. A. Tamor and W. C. Vassel, J. Appl. Phys. 76, 3823 (1994). A. C. Ferrari and J. Robertson, Phys. Rev. B 61, 14095 (2000). C. Lenardi, P. Piseri, V. Briois, A. Li Bassi, C. E. Bottani, and P. Milani, J. Appl. Phys. 85, 7159 (1999). E. Riedo, E. Magnano, S. Rubini, M. Sancrotti, E. Barborini, P. Piseri, and P. Milani, Solid State Commun. 116, 287 (2000). M. Grimsditch, J. Phys. C 1, L143 (1983). X. Jiang, Phys. Rev. B 43, 2372 (1990). X. Jiang, J. W. Zou, K. Reichelt, and P. Gruenberg, J. Appl. Phys. 66, 4729 (1989). D. Fioretto, G. Carlotti, G. Socino, S. Modesti, C. Cepek, L. Giovannini, O. Donzelli, and F. Nizzoli, Phys. Rev. B 52, R8707 (1995). M. Manfredini, C. E. Bottani, and P. Milani, Chem. Phys. Lett. 226, 600 (1994). P. Milani, A. Podestà, P. Piseri, E. Barborini, C. Lenardi, and C. Castelnovo, Diamond Rel. Mater. 10, 240 (2001). A. Pinczuk and E. Burstein, in “Light Scattering in Solids” (M. Cardona, Ed.), Topics in Applied Physics. Springer, Berlin, 1975. B. Lengeler, in “X Ray Absorption and Reflection in the Hard X-Ray Range” (M. Campagna and K. Rosei, Eds.). NorthHolland, Amsterdam, 1990. C. Casari, A. Li Bassi, C. E. Bottani, E. Barborini, A. Podestà, and P. Piseri, Diamond Rel. Mater. 12, 856 (2003). D. A. Glocker and S. I. Shah, Eds., “Handbook of Thin Film Process Technology.” IoP, Bristol, 1995.

Cluster-Assembled Nanostructured Carbon 78. F. Siviero, E. Barborini, C. Boragno, R. Buzio, E. Gnecco, C. Lenardi, P. Piseri, S. Vinati, U. Valbusa, and P. Milani, Surf. Sci. 513, 381 (2002). 79. K. H. Guenther, Appl. Opt. 23, 3806 (1984). 80. M. Chowalla, R. A. Aharonov, C. J. Kiely, I. Alexandrou, and G. A. J. Amaratunga, Phil. Mag. Lett. 75, 329 (1997). 81. J. Schäfer, J. Ristein, R. Graupner, L. Ley, U. Stephan, Th. Frauenheim, V. S. Veerasamy, G. A. J. Amaratunga, M. Weiler, and H. Ehrhardt, Phys. Rev. B 53, 7762 (1996). 82. J. J. Yeh and I. Lindau, Atomic Data Nucl. Tables 32, 1 (1985). 83. P. Kovarik, E. B. D. Bourdon, and R. H. Prince, Phys. Rev. B 48, 12123 (1993). 84. E. Magnano, C. Cepek, M. Sancrotti, F. Siviero, S. Vinati, C. Lenardi, P. Piseri, E. Barborini, and P. Milani, Phys. Rev. B 67, 125414 (2003). 85. P. E. Batson, Phys. Rev. B 48, 2608 (1993). 86. M. Jaouen, G. Tourillon, J. Delafond, N. Junqua, and G. Hug, Diamond Rel. Mater. 4, 200 (1995). 87. J. F. Morar, F. J. Himpsel, G. Hollinger, G. Hughes, and J. L. Jordan, Phys. Rev. Lett. 54, 1960 (1985). 88. J. Robertson, Philos. Mag. Lett. 57, 143 (1988). 89. V. Paillard, M. Meaudre, P. Melinon, V. Dupuis, J. P. Perez, S. Vignoli, A. Perez, and R. Meaudre, J. Non-Cryst. Solids 191, 174 (1995). 90. M. Bruzzi, P. Piseri, S. Miglio, G. Bongiorno, E. Barborini, C. Ducati, J. Robertson, and P. Milani, Euro. Phys. J. B, in press. 91. N. F. Mott and E. A. Davis, “Electronic processes in noncrystalline materials”, Oxford University Press, Oxford, 1979. 92. K. Shimakawa and K. Miyake Phys. Rev. B 39, 7578 (1989). 93. J. Robertson, J. P. Sullivan, O. Zhou, T. B. Allen, and B. F. Coll, “Amorphous and Nanostructured Carbon.” MRS Symp. Proc., Vol. 593, 2000. 94. G. A. J. Amaratunga and S. R. P. Silva, J. Non-Cryst. Solids 198, 611 (1996). 95. B. S. Satyanarayana, A. Hart, W. I. Milne, and J. Robertson, Appl. Phys. Lett. 71, 1430 (1997). 96. O Groening, O. M. Kuettel, P. Groening, and L. Schlapbach, Appl. Phys. Lett. 71, 2253 (1997). 97. J. Robertson, J. Vac. Sci. Technol. B 17, 695 (1999). 98. W. A. deHeer, A. Chatelain, and D. Ugarte, Science 270, 1179 (1995). 99. O. Gröning, O. M. Küttel, Ch. Emmenegger, P. Gröning, and L. Schlapbach, and J. Vac. Sci. Technol. B 18, 665 (2000). 100. W. Zhu, G. P. Kochanski, and S. Jin, Science 282, 1471 (1998). 101. A. C. Ferrari, B. S. Satyanarayana, J. Robertson, W. I. Milne, E. Barborini, P. Piseri, and P. Milani, Europhys. Lett. 46, 245 (1999). 102. A. Ilie, A. C. Ferrari, T. Yagi, S. E. Rodil, J. Robertson, E. Barborini, and P. Milani, J. Appl. Phys. 90, 2024 (2001). 103. C. Ducati, E. Barborini, P. Piseri, P. Milani, and J. Robertson, J. Appl. Phys. 92, 5482 (2002). 104. T. W. Mercer, N. J. DiNardo, J. B. Rothman, M. P. Siegal, T. A. Friedman, and L. J. Martinez-Miranda, Appl. Phys. Lett. 72, 2244 (1998). 105. O. L. Blakslee, D. G. Proctor, E. J. Seldin, G. B. Spence, and T. Weng, J. Appl. Phys. 41, 3373 (1970). 106. E. J. Seldin and C. W. Nezbeda, J. Appl. Phys. 41, 3389 (1970). 107. B. Bhushan, “Handbook of Micro and Nano Tribology.” Chemical Rubber, Boca Raton, FL, 1995. 108. R. Carpick and M. Salmeron, Chem. Rev. 97, 1163 (1997). 109. MEMS Issue, Mater. Res. Soc. Bull. 26 (2001), pp. 282–337. 110. J. Sullivan, T. Friedmann, and K. Hjort, Mater. Res. Soc. Bull. 26, 309 (2001). 111. S. Sundararajan and B. Bhushan, Wear 225/229, 678 (1999). 112. B. Bhushan, Trans. ASME, J. Tribol. 116, 378 (1994). 113. G. Dedkov, Phys. Stat. Sol. 0a 179, 3 (2000).

25 114. S. Sundararajan and B. Bhushan, J. Appl. Phys. 88, 4825 (2000). 115. V. Koinkar and B. Bhushan, J. Appl. Phys. 81, 2472 (1997). 116. A. Podestà, G. Fantoni, and P. Milani, Quantitative nanofriction characterization of corrugated surfaces by atomic force microscopy (unpublished). 117. A. Podestà, G. Fantoni, P. Milani, M. Ragazzi, D. Donadio, and L. Colombo, J. Nanosci. Nanotechnol. 2, (2002). 118. A. Grill, Surf. Coat. Technol. 94/95, 507 (1997). 119. M. Binggeli and C. Mate, Appl. Phys. Lett. 65, 415 (1994). 120. V. Koinkar and B. Bhushan, J. Vac. Sci. Technol. A 14, 2378 (1996). 121. B. Bhushan and C. Dandavate, J. Appl. Phys. 87, 1201 (2000). 122. L. Ostrovskaya, A. Podestà, P. Milani, and V. Ralchenko, Europhys. Lett. 63, 401 (2003). 123. E. Gnecco, R. Buzio, C. Boragno, and U. Valbusa, Acta Phys. Slovaca 50, 423 (2000). 124. R. Buzio, E. Gnecco, C. Boragno, and U. Valbusa, Carbon 40, 883 (2002). 125. C. Putman and M. Igarashi, Appl. Phys. Lett. 66 (1995). 126. L. Schlapbach and A. Zuettel, Nature 414, 353 (2001). 127. S. Hynek, W. Fuller, and J. Bentley, Int. J. Hydrogen Energy 22, 601 (1997). 128. M. S. Dresselhaus, K. A. Williams, and P. C. Eklund, MRS Bull. 45 (1999). 129. A. C. Dillon, K. M. Jones, T. A. Bekkedahl, C. H. Kiang, D. S. Bethune, and M. J. Heben, Nature 386, 377 (1997). 130. C. Liu, Y. Y. Fan, M. Liu, H. T. Cong, H. M. Cheng, and M. S. Dresselhaus, Science 286, 1127 (1999). 131. Y. Ye, C. C. Ahn, C. Witham, B. Fultz, J. Liu, A. G. Rinzler, D. Colbert, K. A. Smith, and R. E. Smalley, Appl. Phys. Lett. 74, 2307 (1999). 132. P. Chen, X. Wu, J. Lin, and K. L. Tan, Science 285, 91 (1999). 133. A. Chambers, C. Park, R. T. K. Baker, and N. M. Rodriguez, J. Phys. Chem. B 102, 4253 (1998). 134. C. Park, P. E. A. A. Chambers, C. D. Tan, R. Hidalgo, and N. M. Rodriguez, J. Phys. Chem. B 103, 10572 (1999). 135. F. Darkrim and D. Levesque, J. Chem. Phys. 109, 4981 (1998). 136. Q. Wang and J. K. Johnson, J. Chem. Phys. 110, 577 (1999). 137. S. M. Lee, K. S. Park, Y. C. Choi, Y. S. Park, J. M. Bok, D. J. Bae, K. S. Nahm, Y. G. Choi, S. C. Yu, N. Kim, T. Frauenheim, and Y. H. Lee, Synth. Met. 113, 209 (2000). 138. S. T. Mayer, R. W. Pekala, and J. L. Kaschmitter, J. Electrochem. Soc. 140, 446 (1993). 139. A. F. Burke, “1st International Seminar on Double Layer Capacitor and Similar Energy Storage Devices,” Florida Educational Seminars, Boca Raton, FL, 1991. 140. A. W. Adamson, “Physical Chemistry of Surfaces.” Wiley, New York, 1976. 141. S. R. Morrison, “The Chemical Physics of Surfaces.” Plenum, New York, 1990. 142. O. Stern, Z. Electrokem. 30, 508 (1924). 143. C. Niu, E. K. Sichel, R. Hoch, and H. Tennent, Appl. Phys. Lett. 70, 1480 (1997). 144. A. Kornyshev, W. Schmickler, and M. Vorotyntsev, Phys. Rev. B 25, 5244 (1982). 145. I. Tanahashi, A. Yoshida, and A. Nishino, J. Electrochem. Soc. 137, 3052 (1990). 146. M. Bruzzi, P. Piseri, E. Barborini, G. Benedek, and P. Milani, Diamond Rel. Mater. 10, 989 (2001). 147. L. Diederich, E. Barborini, P. Piseri, A. Podestà, P. Milani, A. Schneuwly, and R. Gallay, Appl. Phys. Lett. 75, 2662 (1999). 148. Y. Mihara, in “Frontiers of Nano-Optoelectronic Systems,” (L. Pavesi and E. Buzaneva Eds.). NATO ASI, Kluwer, Dordrecht, 275, 2000. 149. S. M. Rossnagel, D. Mikalsen, H. Kinoshita, and J. J. Cuomo, J. Vac. Sci. Technol. A9 261 (1991). 150. Y. Xia and G. M. Whitesides, Angew. Chem., Int. Ed. 37, 550 (1998).

26 151. D. C. Liu and C. P. Lee, Appl. Phys. Lett. 63, 3503 (1993). 152. P. Gatz and O. F. Hagena, Appl. Surf. Sci. 91, 169 (1995). 153. E. Barborini, P. Piseri, A. Podestà, and P. Milani, Appl. Phys. Lett. 77, 1059 (2000). 154. A. A. Zakhidov, R. H. Baughman, Z. Iqbal, C. Cui, I. Khayrullin, S. O. Dantas, J. Marti, and V. G. Ralchenko, Science 282, 897 (1998).

Cluster-Assembled Nanostructured Carbon 155. E. Franckowiac, S. Gautier, H. Gaucher, S. Bonnamy, and F. Beguin, Carbon 37, 61 (1999). 156. I. Boscolo, P. Milani, M. Parisotto, G. Benedek, and F. Tazzioli, J. Appl. Phys. 87, 4005 (2000). 157. B. E. Conway, “Electrochemical Supercapacitors: Scientific Fundamentals and Technological Applications.” Kluwer, New York, 1999.

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Colloidal Gold Victor H. Perez-Luna, Kadir Aslan, Pravin Betala Illinois Institute of Technology, Chicago, Illinois, USA

CONTENTS 1. Introduction 2. Preparation Methods 3. Surface Modification and Functionalization of Gold Nanoparticles 4. Characterization 5. Applications 6. Conclusions Glossary References

1. INTRODUCTION There is currently enormous interest in the field of nanoscale science and engineering because of (i) the novel properties of matter found at the intermediate scale between molecules and bulk materials and (ii) phenomena that occur only at the nanoscale regime. Over the last decade the number of publications in the field of nanoparticle synthesis has increased enormously. Gold nanoparticles in particular are receiving renewed interest because of novel applications being developed with this material. One of the reasons for this renewed interest is the attractive combination of its optical properties, which depend on the electronic properties rather than molecular structure, and the facile functionalization of this material with a variety of molecules. Despite being known and used for centuries, the use of this material is currently exhibiting a boom, particularly in the field of materials science.

1.1. History Accounts on the use of colloidal gold can be found throughout history. The oldest object containing gold nanoparticles is frequently cited as being the Lycurgus Chalice from fifth century Rome [1, 2]. However, accounts of colloidal gold preparations may extend further back in time. Colloidal gold is found in treatments in the alternative medicine Ayurveda

ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

(the traditional medical system of India), which has its written origins in the Vedas, the sacred texts of India. The Vedas are believed to be the oldest writings in the world. More recent accounts of colloidal gold are its uses in alchemia by Paracelsus around 1600 [3], incorporation of gold into molten silica to create “ruby glass” [attributed to Johann Kunckel (1638–1703)] [4, 5], and the experiments by Michael Faraday that led him to attribute the color of ruby glass and his aqueous solutions of gold (colloidal gold) to finely divided gold particles [6, 7]. In addition to the early experiments by Faraday, colloidal gold was considered a useful research tool for particle science in the early 19th century. This led to important advances in the development of novel synthetic methods for colloidal gold [8]. However, its main applications were still relegated to the production of dyes for glass and fabrics. The development of a reproducible method through the sodium citrate reduction method to produce colloidal gold with controlled and reproducible properties [9] eventually made possible its application to the fields of particle science and cell biology [10–14]. Until recently, the wide use of colloidal gold for labeling studies in cell biology was one of the more important applications of this material with a good number of textbooks devoted to the subject [10–14]. However, during the last few years there has been a resurgence of interest in colloidal gold because of the potential of this material for a variety of applications related to nanoscale science and engineering. The scope of this chapter will concentrate on these latest developments, particularly in their relationship to nanoscale science and engineering.

1.2. Properties Due to its large free electron density, colloidal gold has been used extensively in applications involving electron microscopy. As it turns out, the free electrons of metallic nanoparticles also impart them with unique optical properties, which can be explained in terms of classical free-electron theory and simple electrostatic models of electrostatic polarizability [7, 14]. It was Michael Faraday who first explained the characteristic colors of colloidal dispersions of gold and ruby glass. Although his theoretical framework was mostly qualitative, it formed the foundations for

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (27–49)

28

Colloidal Gold

with 0 being the dielectric constant of vacuum, m the dielectric constant of the media the nanoparticles are embedded into, and  the dielectric constant of the particle (assumed equal to that of the bulk metal). For metals, the dielectric constant is a complex number related to the index of refraction n and absorption coefficient k as  = 1   + i2   = n + ik2

(2)

and the magnitude of 1 and 2 are strong functions of the frequency of the applied electric field (wavelength of the incident light). A direct result of Eqs. (1) and (2) is that the polarizability of a metallic nanoparticle will be maximum when the term  + 2m is minimum, that is, when 1 + 2m 2 + 22 = minimum. During a typical ultraviolet-visible measurement the extinction coefficient  is related to k as  = 4k/. The overall absorption was obtained first by Mie [19] by integrating over a medium filled with nanoparticles according to 3 k = 9/23/2 m N 4/3R 

2 1 + 2m 2 + 22

(3)

where N is the number density of the nanoparticles. The term in parentheses represents the filling factor f = N 4/3R3 . For dilute dispersions, Eq. (3) describes the absorption of noninteracting particles. However, when the filling factor is large, each particle will be subjected to an average polarization field due to surrounding particles. Under these conditions, it is helpful to calculate an effective dielectric constant 1 + 2f 1 + 2N/30 m  = m eff = m 1 − N/30 m  1 − f   − m = =  + 2m 30 m V

with (4)

Equation (4) is known as the Maxwell–Garnett formula. A direct consequence of the static approximation is that the effective optical properties are independent of particle size. The imaginary part of the refractive index is     k=

05 −eff 1 +

2eff 1 + 2eff 2

(5)

In order to calculate an extinction coefficient it is necessary to first obtain an estimate of the dielectric constants. For metals, the dielectric constants can be approximated by the Drude model [20]. According to this model the real and imaginary parts of the dielectric constant are given by 1 = 1 −

2p

and

2 = 1 −

2p 3

d

(6)

 where p = ne2 /0 me is the bulk plasma frequency, n is the concentration of free electrons in the metal, and m and e are the effective mass and charge of the electron respectively. The term d = vF /l represents the relaxation or damping frequency with vF being the velocity of the electrons at the Fermi level and l the electron mean free path [14–17, 20, 21]. The Drude model allows an estimation of the dielectric constants. In addition, tabulated values of experimentally determined dielectric constants for bulk metals can be found in the literature [22]. Figure 1 shows the simulated extinction coefficient of 10 nm gold nanoparticles embedded in media with different dielectric constants. Tabulated experimental values of the dielectric constants of gold were used to simulate the spectra [22]. This figure shows the sensitivity of the absorbance of the nanoparticles to the dielectric medium surrounding them. Two features that emerge from this figure are a slight redshift in the surface plasmon resonance (SPR) peak position and an increased intensity of the SPR absorption band. Thus, changing the dielectric constant of the media surrounding the nanoparticles results in changes on their optical absorption spectra. Interactions of other materials with the nanoparticles’ surfaces can induce similar changes. The optical properties of gold nanoparticles are also susceptible to the degree of interaction (aggregation) among the particles. For a particle pair with a small enough centerto-center separation, D, the oscillations of the surface plasmons differ significantly from those of isolated particles. In aggregated pairs, two excitation modes can exist, parallel or perpendicular relative to the axis pair. Longitudinal inphase oscillations differ significantly from transversal oscillations with the former being significantly different from noninteracting particles while the latter is only slightly redshifted. When interacting nanoparticle pairs are exposed to

0.06 0.05

extinction

more elaborate theories on the colors of metallic colloidal dispersions. The optical properties of metallic nanoparticles with radius, R, much smaller than the wavelength, , of light 2R/ ≪ 1 can be estimated by assuming that the size dependent interaction of the particles with light is static (a direct result of 2R/ ≪ 1 while the dielectric properties of the free electrons in metallic nanoparticles are considered dynamic. Under these assumptions, the theory predicting the optical properties of metallic clusters is represented by the sets of equations known as the Maxwell-Garnett formula [14–18]. The polarizability  of a small spherical particle in a static electric field is  − m (1)  = 40 m R3  + 2m

0.04 0.03 0.02 0.01 0.00 400

500

600

700

800

λ (nm) Figure 1. Simulated optical absorption spectra for 10 nm gold particles embedded in media with m = 10 177 20, and 25.

29

Colloidal Gold

an applied electrical field E, the resulting dipole moments, 1j and 2j , of the two spheres are given by 1j = 1 E ∗ +

1 2j j 40 m D3

2j = 2 E ∗ +

2 1j j 40 m D3

and (7)

with j = 1 representing the transversal mode and j = 2 the longitudinal mode of oscillation. If the angle between the electric field vector and the longitudinal axis of the particle pair is , then E ∗ = E cos  for the longitudinal mode and E ∗ = E sin  for the transversal mode. The pair polarizability for the longitudinal and transversal modes can be calculated with these equations. For nanoparticle pairs in a two-dimensional (2D) plane, the average polarizability pair 2D is obtained when integrating over all possible orientations of the longitudinal and transversal modes, giving pair 2D =

2

1 − 2 R/D3 −1 + 1 + R/D3 −1 

(8)

where =  − m  + 2m . With the average polarizability pair 2D , an effective dielectric constant and the absorption can be calculated with Eqs. (4) and (5), respectively. Figure 2 shows the simulated absorption spectra of 10 nm gold nanoparticles embedded in dielectric medium with m = 25 at various separation distances between the nanoparticles. When the particles are separated by a large distance they behave as independent particles and the simulated spectrum is similar to that obtained from Eq. (3). However, when the separation becomes comparable to the radius of the particle, longitudinal and transversal resonance peaks can coexist. The longitudinal mode is significantly redshifted while the transversal mode is only slightly redshifted from that of noninteracting particles. Furthermore, the magnitude and shift of the longitudinal mode become significant when the separation distance is smaller than the diameter of the particles and this effect becomes stronger as the separation distance of the particles becomes smaller. The previous approximation is valid for 2D particle-pair aggregates. For many particle aggregates, more complex calculations are

extinction

0.08 0.06 0.04 0.02 0.00 400

500

600

700

800

ν (nm) Figure 2. Simulated optical absorption spectra of 10 nm gold nanoparticles at varying separation distances between the particles with R/D = 00001, 0.444, 2.5, and 2.0.

necessary [23, 24]. In this case, additional resonances appear at lower frequencies (longer wavelengths) that are even more separated than those calculated for 2D particle pairs. In fact, their optical absorption spectra can show broad bands that extend into the near infrared region depending on the size of the aggregates and the number of particles in the aggregate [17, 18, 23–25]. Because particle size, degree of aggregation, and the dielectric constant of the media determine its optical properties, applications of colloidal gold exploiting the strong dependence of optical properties with aggregation and surrounding dielectric media have been implemented through proper surface modification of the nanoparticles or by embedding them in different dielectric media. The following sections summarize preparation methods, surface modification, characterization, and modern applications. Traditional applications (e.g., immunostaining) are only briefly discussed since there is already a wide body of literature on these applications and the techniques are well established [10–14].

2. PREPARATION METHODS 2.1. Gold Nanoparticle/Colloidal Gold Synthesis Metallic nanoparticles can be synthesized by a variety of methods such as chemical [10, 12–14, 17], sonolytic [26], radiolytic [27–30], and photolytic reactions [31–37]. However, the reduction of metal salts by added reducing agents is the oldest established procedure for the preparation of aqueous suspensions of nanoparticles. A wide range of reducing agents have been used in gold nanoparticle synthesis; the most frequently used ones include sodium citrate [38, 39], sodium borohydride [10, 12–14, 40], tannic acid [10, 12–14], white phosphorus [6, 10, 41] hydrazine [42–44], hydrogen [45, 46], carbon monoxide [47], hydroxylamine [48–50], and mild agents like ascorbic acid [51, 52]. Reducing solvents such as alcohols [53–55], formamide [56], and ethers [54, 57] can also be used in the synthesis of metallic nanoparticles. In 1951, Turkevich et al. [58]. established the first reliable procedure for the preparation of gold sols with wellcontrolled size distributions. Such a method is based on reduction of chloroauric acid by trisodium citrate and has become a standard for histological staining applications. Other approaches to the synthesis of colloidal gold nanoparticles have been used. Among those is a two-phase system, first introduced by Faraday [6], who reduced aqueous gold salt with phosphorus vapor in carbon disulfide. Later, a combination of the two-phase approach with the more recent techniques of ion extraction and formation of self-assembled monolayers (SAMs) of alkanethiols on gold led to a one-step method for the synthesis of derivatized gold nanoparticles by Brust et al. [59, 60]. The technique consists of simultaneous reduction of gold salts and SAM formation. The redox reactions can be carried out by an appropriate choice of reducing and phase-transfer reagents in the two phases. In the original procedure, an aqueous solution of chloroaurate is mixed with a solution of the phase-transfer reagent tetraoctyl ammonium bromide in

30 toluene and the mixture is vigorously stirred until all tetrachloroaurate was transferred to the organic layer. The alkanethiol is then added to the organic phase followed by the reducing agent. This procedure led to the formation of stable gold nanoparticles protected by a self-assembled monolayer that imparted the nanoparticles different properties depending on the functional group at the surface. These latter systems are also commonly referred to as monolayer protected clusters, which have the characteristic that they can be treated as unique chemical entities because they can be manipulated with a variety of physicochemical procedures while maintaining their identity (vide infra) [59–61].

2.2. Polymers in the Synthesis of Gold Nanoparticles The preparation of nanocomposites with an inorganic core such as metal nanoparticles and a protective organic shell of small molar mass compounds or polymers is a route to obtain materials suitable for hierarchical self-organization into nanoassemblies. Incorporation of nanoparticles into polymer matrices is particularly advantageous for materials engineering and the study of nanoparticle–matrix interactions. The adsorption of polymers onto a wide variety of nanoparticles is of broad industrial relevance in such areas as pharmaceuticals, photography, paints and coatings, foods, wastewater treatment, rubbers, and composite materials and is ubiquitous in natural aqueous environments as well as in many biological organisms. Recently, nanocomposites of -conjugated polymers and inorganic particles have received much attention [62–68], due to the interesting high electrical conductivity [69] and redox properties [70] of -conjugated polymers and their extensive applications ranging from batteries to light-emitting devices [71]. The adsorption or grafting of environmentally responsive “smart” polymers onto surfaces holds potential applications in controlled modulation of interfacial properties, as demonstrated in thermally switchable cell culture of anchorage-dependent mammalian cells [72, 73] and to control bacteria fouling [74]. Considering the wide range of applications, the combination of polymers with nanoparticles has driven lot of interest in its research. Overall there have been two different approaches to achieve nanoparticle–polymer composites. The first technique consists of the in-situ preparation of the nanoparticles in the matrix. This is effected either by the reduction of metal salts dissolved in the polymer matrix [75–77] or by the evaporation of metals on the heated polymer surface [78]. A second technique involves polymerizing the matrix around the nanoparticles [79]. Lennox et al. [80]. proposed a desirable approach that involved blending premade nanoparticles into a presynthesized polymer matrix. Coating gold nanoparticles first with a polymer ligand chemically similar to the polymer matrix made mixing of the gold nanoparticles with the polymeric matrix thermodynamically more favorable and thus more compatible as compared to coatings with polymer ligands that were chemically different from the matrix [81]. The surfaces of gold nanoparticles were decorated with covalently bound thiol-capped polystyrene macromolecules (PS–SH) and these were dispersed in a PS matrix. Because of its

Colloidal Gold

high glass-transition temperature, PS allows for the formation of robust films at room temperature. The PS–SH macromolecules were synthesized by anionic polymerization and the polystyrene anion was titrated with one unit of propylene sulfide to generate the thiol end group [82]. The PS–Au nanoparticles were prepared following modified Ulman reaction conditions [83], using lithium triethylborohydride as a reducing agent. Incorporation into the PS matrix was effected by dissolving both PS–Au and PS matrix in a small amount of chloroform. The PS–Au nanoparticles were found to be evenly dispersed. This method provided for the full synthetic control over both the nanoparticles and the matrix. Novel -conjugated polymer-protected metal (Pd, Au, Pt) nanoparticles have been prepared via reduction of the metal ions by a -conjugated poly(dithiafulvene) (PDF) having strong electron-donating properties. For the synthesis of gold nanoparticles, the PDF is dissolved into a dimethyl sulfoxide solution of HAuCl4 and the reaction mixture is stirred for 24 hours at room temperature. The reaction mixture changes gradually from yellow to purple, indicating nanoparticle formation. It has been proposed that electron transfer from the -conjugated PDF to the Au ions results in the formation of Au nanoparticles embedded in the polymeric matrix. The oxidized -conjugated polymer might possess a delocalized positive charge and provide both steric and electrostatic stabilization, protecting the Au in stable colloidal form [84]. Nanoparticles encapsulated in -conjugated polymers show interesting characteristics, particularly in dielectric properties, energy storage, catalytic activity, and magnetic susceptibility [85]. Hybrid polymeric systems with metal nanoparticles located inside specific areas of the material have been developed by Svergun et al. [86]. In this approach, a selfassembled multilayered polymer, poly(octadecylsiloxane), provided a nanostructured matrix for the incorporation of gold and platinum compounds and for metal nanoparticle formation. The size distributions of the nanoparticles and the changes in the internal structure of the matrix as a result of nanoparticle formation for different reducing agents and reaction media were studied. In addition to the examples mentioned, a variety of other polymers used for the preparation and study of polymer– gold nanocomposites can be found in the literature [87–91]. Further uses of polymers also involve surface modification of nanoparticles as described in Section 3.5.

2.3. Laser Ablation In recent years, gold nanoparticles with diameters less than 10 nm have attracted much attention because of their size dependent physical and chemical properties: a gold nanoparticle in the size range of less than 5 nm in diameter shows a dramatic decrease of the melting point [92, 93], exhibits intense photoluminescence [94], and catalyzes CO oxidation even at low temperatures (200 K) when used with TiO2 [95, 96]. Traditionally, chemical reduction of gold salt into gold nanoparticles in reverse micelles has been one of the most successful methods to obtain gold nanoparticles with diameters less than 10 nm in solution [97]. In this method, the size of the nanoparticles depends on that of the

31

Colloidal Gold

reversed micelles, so that nanoparticles with a desired size can be obtained by introducing a proper amount of water inside the reverse micelles through adjustment of the molar ratio of water to nonpolar solvent. Alternatives to chemical synthesis for the preparation of gold nanoparticles with diameters less than 10 nm have been developed that involve laser ablation of a gold plate in solution [98]. These methods are currently under investigation by several research groups [99–102]. In this method, a gold metal plate, which is placed at the bottom of a glass vessel filled with an aqueous solution of a surfactant (e.g., sodium dodecyl sulfate or SDS), is irradiated with a Nd:YAG pulsed laser (1064 nm) and gold nanoparticles are formed [101]. Formation of gold nanoparticles occurs by a SDS concentration dependent mechanism [103]. After laser ablation, a dense cloud of gold atoms is built over the laser spot of the metal plate and the atoms aggregate as fast as they are supplied. This continues until the atoms in close vicinity are depleted almost completely. As a result an embryonic particle is formed in a region void of atoms. However, the supply of atoms outside the region by diffusion causes the particle to grow slowly even after rapid growth ceases. The surfactant (SDS) plays an important role determining the stability and size of the nanoparticles, because the termination of the particle growth is controlled by the diffusion and attachment rates of SDS on the nanoparticle. Gold nanoparticles end up coated with a double layer of SDS, such that in the first layer the SDS molecules orient with the hydrophilic −SO− 4 end toward the nanoparticle and the hydrophobic −C12 H25 end outward while in the second layer they orient in the opposite direction. The fact that the nanoparticles end up coated with a double layer of SDS ensures the stability of gold nanoparticles in solution. Gold nanoparticles produced upon laser irradiation (1064 nm) to the gold plate have typical size distributions of 1–15 nm with average diameters less than 10 nm. However, when the nanoparticles form, it is very difficult to change their size, and the aggregation of the ablated atoms can hardly be controlled. In order to address this problem, further size reduction of the gold nanoparticles can be induced by laser irradiation (532 nm) in the vicinity of the wavelength of SPR excitation of the gold nanoparticles (520 nm) [98, 104, 105]. Gold nanoparticles in solution do not absorb photons at 1064 nm, but they absorb photons at 532 nm and fragment into smaller particles. Dissociation of the gold nanoparticles by absorption of 532 nm photons can be achieved by first preparing nanoparticles by laser ablation of the metal at 1064 nm followed by irradiation with a 532 nm laser. The mechanism of laser-induced size reduction along with experimental evidence of the hypothesis was proposed by Takami et al. [106]. In this scheme, when the laser (532 nm) irradiates the gold nanoparticles, the plasmons in the gold nanoparticles absorb photons and the electrons are excited. The temperature of the nanoparticles rises above the melting point such that they melt and become liquid. Further size reduction takes place due to vaporization of the gold nanoparticles when the temperature rises to the boiling point such that atoms and/or small particles are ejected via vaporization. The amount of ejected atoms depends on the absorbed laser energy. The maximum diameter is determined by the equilibrium between

the ejected atoms and the deposition of ejected atoms back onto the nanoparticle surface. Because the ejected atoms are very unstable in aqueous solution, they tend to agglomerate and/or deposit on the gold nanoparticles remaining in the solution. Laser-induced size reduction of gold nanoparticles that are produced by laser ablation of gold plates is a promising new technique for the preparation of size-selected nanoparticles with diameters less than 10 nm in solution. This technique is very simple, is versatile compared to other preparation techniques, and can be applied to other metals.

2.4. Ultrasonic Irradiation In the preceding sections, stabilizers were used to control the particle nucleation and growth processes along with parameters such as the initial concentration of the reactants. Although there is an enormous amount of research being conducted with nanoparticles prepared by these methods, there are only few reports elucidating the relationship between the rates of reduction of metal ions and the size of the formed gold nanoparticles due to experimental difficulties in solution systems. Some of the difficulties include (a) the precise determination of the rate of reduction of metal ions because the reaction is often too fast to track the course of the reduction experimentally, and (b) the quantitative analysis for the concentration of unreduced metal ions that are still present in the reaction medium. In recent years the formation of gold nanoparticles in an ultrasound field has been studied as an alternative to other methods [98, 107–109]. Caruso et al. [98]. summarizes the description of the ultrasonic irradiation phenomenon and the details of this method can be found in the references therein. The mechanisms of formation of gold nanoparticles have also been proposed before. [106, 107] In short, ultrasonic irradiation causes pyrolysis of water and other organic compounds present in an aqueous solution resulting in formation of free radicals at extremely high temperatures and pressures. These radicals reduce the gold (III) ions into gold (II), gold (I), and finally gold (0). The reduction of gold (III) with sonochemically formed radical species can be followed by a colorimetric method using NaBr. The rate of gold (III) reduction and the size of the gold nanoparticles can be controlled by the ultrasound irradiation conditions such as temperature, intensity of the ultrasound, and positioning of the reactor. However, this method yields polydisperse gold nanoparticles, which is an important problem for some applications where monodisperse solutions of gold nanoparticles are required.

2.5. Anisometric Gold Nanoparticles The preparation and characterization of nanostructured materials has grown constantly because of their distinctive properties and potential use in technological applications. Since the electronic, magnetic, catalytic, and optical properties of metallic nanoparticles depend mainly on their size and shape (or morphology), shape control is one of the alternative tools to adjust these properties for various applications.

32 Nonspherical gold nanoparticles in the shapes of platelike triangles [58, 110, 111], rods [112, 113], oblate or pancakelike [114–117], prolate or needlelike [114–117], hexagonal [110], ellipsoidal, quasi-ellipsoidal [110], and noncentrosymmetric [118] have been prepared and studied by a number of groups around the world. Some of the techniques that have been exploited to prepare shape controlled gold nanoparticles include the template synthesis method [115–117, 119–125], ultraviolet (UV) irradiation photoreduction [110], electrochemical synthesis [126], poly(tetrafluoroethylene) (PTFE) friction transfer approach [127], and grazing incidence vacuum deposition [128, 129]. In the template method, different materials are used as templates for the synthesis of anisometric gold nanoparticles. Schönenberger et al. [130]. reported the synthesis of Au nanowires using polycarbonate track-etched membranes as templates. Martin and co-workers [114–117, 119–122, 125] have explored the properties of gold nanowires prepared by electrochemically depositing Au within the pores of alumina membranes. Noncentrosymmetric gold nanoparticle structures with diameters of 26–28 nm were prepared in porous anodic aluminum films via a modified template synthesis procedure [118]. Figure 3 shows the schematic preparation of nanoparticles in an alumina template. The electrochemical preparation of nanoporous alumina films utilized a twoelectrode temperature-controlled anodization cell with lead foil serving as the cathode and the polished aluminum plate as the anode. With constant stirring of the cell, anodic aluminum oxide film was grown to produce pores with 32 nm diameter and film thickness of 40–60 m. Separation of the alumina film was achieved using a voltage reduction procedure, which created small branched pores in the barrier layer of oxide film. After rinsing and drying, the barrier side of the film was sputtered with silver and gold was potentiostatically

Figure 3. Schematic of nanoparticle synthesis: (1) porous anodic alumina film after detachment from Al substrate; (2) one face of alumina film coated with silver using plasma deposition; (3) electrodeposit silver foundation into pores; (4c) electrodeposit gold onto silver foundation to form centrosymmetric gold particles; (4n) electrodeposit layers of gold, silver, and then a second gold layer to form noncentrosymmetric gold particles; (5c and 5n) immerse film in nitric acid to remove silver foundation and silver spacing segments. Reprinted with permission from [118], M. L Sandrock et al., J. Phys. Chem. B. 103, 2668 (1999). © 1999, American Chemical Society.

Colloidal Gold

deposited on a silver foundation to produce centrosymmetric and noncentrosymmetric gold nanoparticles [118]. Esumi et al. [112]. used hexadecyltrimethylammonium chloride (HTAC) (25–30 wt%) as a template for an ultraviolet irradiation technique to synthesize colloidal Au particles with morphologies ranging from spherical to needle shapes. For a 25% (w/w) HTAC concentration, spherical particles were obtained for HAuCl4 concentration below 1 mM and rodlike gold particles at concentrations 5 mM and above. It was discussed that when an aqueous solution of HAuCl4 is added to an aqueous solution of a cationic surfactant, AuCl− 4 forms a 1!1 complex with the cationic surfactant, resulting in a precipitate. With a further increase of the surfactant concentration, the precipitate is redispersed and the solution becomes clear. This situation is understood by the view that AuCl− 4 binds preferentially to the surface of rodlike micelles because of the hydrophilic property of AuCl− 4 and the neutralization of micellar charge. It seems likely that binding of AuCl− 4 to the surface of rodlike micelles reduced by UV irradiation results in formation of rodlike gold particles. In the procedure developed by Chen et al. [110], an aqueous solution of HAuCl4 is mixed with polyvinyl alcohol (PVA), which is used as capping material and irradiated for 48 h, resulting in platelike triangular Au nanoparticles, which developed into hexagonal shapes of about 25 nm with increased HAuCl4 concentration. Also the increase in the duration of irradiation resulted in the formation of more regular platelet triangular shapes. It was reasoned that there was enough time for the reduction of HAuCl4 into elemental Au, which aggregated and grew into more regular triangular shapes. But further study revealed that the crystal morphology of Au nanoparticles was predominantly hexagonal rather than triangular. The variation in Au morphology might be the result of excess gold in the solution, which facilitates the formation of hexagonal Au nanoparticles. The concentration of the capping material played an important role in the formation and morphology of Au nanoparticles. The polymer capping materials acted as molecularly dissolved surface modifiers and steric stabilizers, which influenced the crystallization of the growing Au nucleus site in a specific manner. For PVA concentrations smaller than 1.0 wt%, spherical particles were obtained and a further increase in PVA concentration was found to favor the formation of polyhedralshaped Au nanoparticles. When polyethylene glycol was used in place of PVA as the capping material, Au nanoparticles with quasi-ellipsoidal morphology were obtained. Gold nanorods with different aspect ratios have also been prepared with the aid of shape-inducing micelles using an electrochemical oxidation/reduction method [126]. Appropriate surfactants were employed as both the supporting electrolyte and the stabilizer for the resulting Au nanoparticles. In this approach, a gold metal plate is used as an anode and a platinum plate is used as a cathode. Both plates are immersed in an electrolytic solution consisting of a cationic surfactant (e.g., hexadecylmethylammonium bromide) and a rod-inducing cosurfactant (e.g., tetraoctylammonium bromide/tetradecylammonium bromide). A controlled-current electrolysis is carried out under constant ultrasonication and controlled temperature, resulting in the formation of cylindrical nanoparticles with a typical transverse diameter of 10 nm. The ratio of the two surfactants controls the length of

33

Colloidal Gold

the gold nanorods formed. Hence the surrounding medium, often referred to as capping material, is of great importance as it prevents aggregation and provides the formation of the desired nanorods [113]. It has been observed, in general, that the particle size and shape are influenced by the concentrations of the capping “micellar” materials. Gold nanorods with very large aspect ratios (up to 18) have been prepared following a seed-mediated growth procedure [131, 132]. Wittman and Smith developed the friction-transfer method [133], wherein a block of PTFE is translated across the surface of a glass slide. As a result the glass surface is coated with highly oriented chains of PTFE, which can in turn orient a wide variety of substances from small molecules to polymers [133, 134] and liquid crystals [135]. Alkanethiol coated gold nanoparticles in organic solvents were introduced onto the polymer substrate and subsequently rubbed and flame treated, resulting in strong linear dichroism in the visible spectrum. The gold structures appeared as chains of spherical and rodlike particles with definite gross alignment. The elongated structures appear to be the result of heat-induced particle agglomeration [127]. Ravaine et al. [136]. demonstrated that gold nanoparticles could be formed upon UV irradiation of Langmuir–Blodgett (LB) films of positively charged amphiphiles deposited from an aqueous HAuCl4 subphase. Different LB templates led to different types of Au particles. Single crystals, irregularly shaped crystals, or multiply twinned particles of different size are formed depending on the nature of the amphiphile used to prepare the LB matrix [137, 138]. Most of the particles prepared by this method are platelike particles with sizes ranging from 20 to 800 nm.

2.5.1. Optical Properties of Anisometric Gold Nanoparticles In addition to the interest in controlling the shape of gold nanoparticles, the optical properties of these particles have attracted many researchers since the early 20th century. The optical properties of spherical gold nanoparticles were briefly discussed in Section 1.2 in terms of the surface plasmon resonance absorption peak. The surface plasmon depends on the size of the nanoparticles and many attempts have been made to experimentally correlate the absorption maximum and plasmon bandwidth to the actual size. Several factors, such as nonhomogeneous size distribution and surrounding media, complicate a strict correlation between the theory and experimental results. The optical properties of anisometric gold particles are also of interest because, in addition to size, the absorption spectra are also strongly dependent on the shape of the nanoparticle. In the case of gold nanorods, the optical absorption spectra become split into two SPR modes: a longitudinal mode along the long axis of the rod and a transverse mode perpendicular to the first mode. The magnitude of these modes is also sensitive to the orientational order of the gold rods. With incident light polarized parallel to the main axis of the particles that are fully aligned, the spectrum displays only the longitudinal resonance. In the case of perpendicular polarization of the incident light, the spectrum displays only transverse resonance [113]. The first theoretical calculations were made by Gans [139], who extended

Mie’s theory [19] to prolate and oblate spheroidal particles averaged over all orientations. The Gans theory predicts that the longitudinal resonance shifts to longer wavelengths with increasing aspect ratio (length to diameter ratio), whereas the transverse resonance shifts only slightly to shorter wavelengths [139]. Experimental evidence supporting this theory was obtained by Skillman et al. [140]. who studied the optical properties of silver ellipsoids embedded in gelatin. They found a shift in the longitudinal resonance at longer wavelengths with increasing aspect ratios of the particles. Figure 4 shows a transmission electron microscopy (TEM) picture of gold nanorods synthesized by electrochemical reduction of AuCl4 H in a micellar solution [113, 141, 142]. The same figure shows the UV-visible (UV-vis) absorption spectrum of this system. The presence of transverse and longitudinal SP modes is evident and points out to the special optical properties of these nanoparticles. In summary, spherical gold nanoparticles differ from gold nanorods not only in shape but also in terms of the optical properties. Spherical gold nanoparticles display only one size dependent absorption maximum, whereas gold rods have two maximums depending on the aspect ratio with the longitudinal mode being more intense and shifted to longer wavelengths.

2.6. Nanoshells Nanoshells are nanoparticles normally composed of a dielectric core (usually silica) coated by a thin metallic film, from which the term nanoshell is derived. Synthesis of these particles usually involves formation of dielectric cores (silica nanoparticles) followed by further attachment of very small colloidal gold particles on their surface that will act as nucleation sites for further reduction of gold chloride so that the dielectric nanoparticles become covered by a continuous metallic film. These colloidal systems have similar optical properties to colloidal gold in the sense that they also exhibit SPR absorption. The optical properties of these materials strongly depend on the size of the dielectric core, the thickness of the metallic thin film, and the surrounding media [143–159]. In contrast to colloidal gold, however, where the SPR peak is weakly dependent on particle size, nanoshells allow tuning the SPR peak across the visible and infrared

Figure 4. Right: Experimental UV-vis absorption spectrum of a gold nanorod sample is referred to as the transverse plasmon resonance, while the one centered at 740 nm is called the longitudinal plasmon absorption. Left: TEM image of the same solution. Reprinted with permission from [142], M. A. El-Sayed, Acc. Chem. Res. 34, 257 (2001). © 2001. American Chemical Society.

34 region over a wavelength range of hundreds of nanometers [143, 144, 147, 148, 151, 152, 156]. Thus, while nanoshells have similar properties to colloidal gold, they also allow flexibility in “tuning” their optical properties. Because of this, nanoshells have been investigated as platforms for surface enhanced raman spectroscopy [146], to decrease the rate of photo oxidation of semiconducting polymers [150], and in controlled drug delivery applications [149, 153]. In the latter, the nanoshells can be conjugated to a thermoresponsive polymer carrying a drug. Since tissue can transmit light in the near infrared region with little attenuation, transmitting infrared light to the site where release of the bioactive substance is desired can induce drug release from the thermoresponsive hydrogel [149]. In addition to the tunable optical properties of gold nanoshells composed of a dielectric core, gold nanoshells consisting of a metallic core have also been recently reported [159]. Gold nanoshells with a silver core show larger sensitivity to changes in the properties of the surrounding medium, which makes them attractive materials for biosensing applications. Additionally, these silver/gold core/shell particles also have the advantage of being tunable in their optical properties, in very much the same way as those formed with a dielectric core. Since the outer shell consists of gold, they can also be functionalized in a way similar to conventional colloidal gold [159].

2.7. Other Methods Apart from the chemical and physical synthesis methods, biological production of gold nanoparticles using microorganisms has also been studied by several research groups utilizing the ability of these living organisms to convert Au3+ ions into gold nanoparticles. Beveridge and co-workers [160–163] have shown that gold nanoparticles readily precipitate in bacterial cells when the cells are incubated with Au3+ ions. Mukherjee et al. [164]. demonstrated that the exposure of fungus, Verticillium sp. (AAT-TS-4), to aqueous AuCl− 4 ions resulted in the reduction of the metal ions and formation of gold nanoparticles (diameter of 20 nm). Although the exact mechanism is not known it is speculated that the gold nanoparticles are formed on both the surface and within the fungal cells (on the cytoplasmic membrane) by means of reductases. According to this speculation, the first step involves the trapping of the AuCl− 4 on the surface of the fungal cells. Then, the gold ions are reduced by enzymes within the cell leading to the aggregation of metal atoms and the formation of gold nanoparticles. TEM images of the cells show the presence of gold nanoparticles on the cytoplasmic membrane, which indicates that some of the metal ions/gold nanoparticles diffuse across the cell wall and are localized on the cytoplasmic membrane. It is possible that the enzymes present on the cytoplasmic membrane also participate in the reduction of gold ions. Following this report the same group [165] demonstrated the extracellular formation of gold nanoparticles using another fungus, Fusarium oxysporum. In this case, the formation of gold nanoparticles is believed to be due to reductases released by the fungus into the solution, which causes the reduction of gold ions in the solution. Although these results are promising, more research has to be done in order

Colloidal Gold

to understand the mechanisms of the formation of gold nanoparticles. Braun et al. [166] have also demonstrated that gold nanoclusters can be formed on specific surface sites of protein molecules in analogy to thin film formation of noble metals on inorganic substrates. In the early stages of thin film growth during vacuum deposition, metals first form nanometer size clusters or islands on the substrate if the condensing atoms are more strongly bound to each other than to the substrate (Volmer–Weber mode) [167]. These adsorbed metal atoms are mobile on the surface until they encounter other mobile atoms and come across special sites that have high adsorption and diffusion energies compared to the neighboring places where they are “trapped” and form clusters or islands. Nucleation theories and experimental evidence demonstrate that these clusters are formed in linear arrays (or so-called “decoration lines”), which are caused by steps on the substrate surface as well as the deviation from the perfect crystal surface leading to preferred nucleation of metal clusters on predetermined sites. These principles are also valid for metal condensation on protein molecules and their crystals. During vacuum condensation of gold on frozen enzyme complexes of lumazine synthase (from B. subtilis), proteasome (from T. acidophilum), and GTP cyclohydrolase (from E. Coli), nanoclusters are formed at specific surface sites (decoration). It was found that, under the experimental conditions outlined by Braun et al. [164], gold decorates the surface areas consisting of polar but uncharged residues. The quality of the decoration pattern can be influenced by the temperature of the substrate as well as by attaching dominant trapping sites onto the surface via chemisorption of small molecules or ions such as a polytungstate complex. The decoration sites on the enzyme complexes can be viewed by electron microscopy and their atomic surface structures can be determined with X-ray crystallography. These results opened ways to analyze the binding mechanism of nanoclusters to small specific sites on the surface of hydrated bio-macromolecules by physical– chemical methods. Although very little is known about the nature of the special sites on protein surfaces and why they enhance or impede cluster formation, understanding the binding mechanism may lead to decoration techniques with much fewer background clusters. This would improve the analysis of single molecules with regard to their symmetries and their orientation in the adsorbed state and in precrystalline assemblies as well as facilitate the detection of point defects in crystals caused by misorientation or by impurities.

3. SURFACE MODIFICATION AND FUNCTIONALIZATION OF GOLD NANOPARTICLES In order to fully take advantage of the unique properties of gold nanoparticles, it is necessary to impart them with different surface functionalities. This allows the nanoparticles to be readily dispersed in nonpolar media and make them more compatible with a matrix during the formation of nanocomposites, or it can provide the nanoparticles with the specific molecular recognition capabilities of biomolecules.

35

Colloidal Gold

A variety of methods for surface modification exist that can be implemented with nanoparticles. Maintaining the stability of the nanoparticles upon surface modification is important. However, gold nanoparticles often tend to aggregate upon surface modification, especially when the surface functionalization proceeds with elimination of the negative charges on colloidal gold that normally keeps these particles from aggregating in aqueous solutions. Such is the case with the formation of self-assembled monolayers of alkanethiolates on colloidal gold [168, 169]. Chemical modification of gold surfaces can be performed by using alkanethiols, polymers, dendrimers, and other molecules as stabilizers. This can be achieved either by surface modification of already available colloidal gold or by the synthesis of colloidal gold with an organic monolayer in a one-step procedure. The following section gives a brief description of surface modification techniques for colloidal gold.

3.1. Monolayer Protected Clusters Chemisorption of alkane thiols offers a facile route for creating well-defined surface chemistries on gold substrates. This area of research has been extensively explored during the last two decades [170–177]. Surface modification of colloidal gold using alkane thiols would appear to be a logical choice to modify the surface chemistry of these nanoparticles. However, chemisorption of alkane thiols eliminates the charges on colloidal gold that impart these particles with their stability in aqueous environments, leading to destabilization and hence irreversible flocculation [168, 169]. It is possible to synthesize stable thiol-functionalized gold nanoparticles in a one-step procedure. Common synthetic procedures to achieve this goal are most commonly based on modifications to the synthetic procedure originally reported by Brust et al. (described briefly in Section 2) [59, 60]. This method results in nanoparticles protected by an organic monolayer, which are commonly referred to as monolayer-protected clusters (MPCs). The alkanethiolate molecules protecting the gold nanoparticles have the potential to develop novel technological applications such as chemical sensing, catalysis, colorimetric assays for DNA detection, biosensors, etc [61, 178]. Understanding the properties of the alkanethiolate ligands may expand the range of applications of these nanoparticles. A wide variety of thiol molecules such as straight-chain alkanethiols [179, 180], glutathione [181], tiopronin [182], thiolated poly(ethylene glycol) [183], p-mercaptophenol [184], aromatic alkane thiol [185], phenyl alkanethiols [186], and -mercaptopropyl)trimethoxysilane [187] have been used to prepare MPCs. Several research groups have undertaken extensive efforts toward the synthesis and characterization of MPCs. Notable among these are the papers of Murray et al. who followed the first reports of Brust et al. [59, 60]. to develop elaborate characterizations of the gold clusters stabilized by monolayers of alkane thiols and a route to mixed monolayers of alkane thiols and -alkane thiols ligands on these nanometer-sized clusters [188]. The monolayers on these nanoparticles confer these systems with new surface properties and with extraordinary stability. In fact, they can be isolated and manipulated with a variety of chemical and physical procedures to the extent that they can be dried

and resuspended in solution again without compromising the stability of MPCs. Subsequent reports by this group have shown that a variety on alkanethiol chain lengths (C3 –C24  [179, 180], -functionalized alkane thiols [189], and dialkyl disulfides [190] can be employed in the same protocol. Several groups in the world are now actively involved in the study and synthesis of these nanoparticles [40, 61, 191–197].

3.2. Place-Exchange Reaction Although the synthesis of a number of alkanethiolate functionalized MPCs is reported in detail in the literature, more diverse surface properties can be achieved by altering the surface with different chemical groups. One method to accomplish further functionalization of MPCs is the place-exchange reaction, which has been used to obtain a wider variety of MPCs by various research groups [61, 188, 198–203]. Noteworthy are the efforts during the last decade of Murray et al. to produce MPCs with a variety of ligands by the place-exchange reaction [198, 199, 202]. In this method, the alkanethiolates of MPCs are displaced by chemisorption of other alkanethiols during incubation in a concentrated solution of alkanethiols. Two routes can be applied toward producing poly-hetero- -functionalized clusters: (a) simultaneous, exchange of a mixture of thiols onto an MPC and (b) stepwise, progressive exchange of different thiols, isolating and characterizing the cluster product after each step. During a typical place-exchange reaction, MPCs with alkanethiolate monolayers (RS) are functionalized with R′ S groups according to xR′ SH + RSm MPC → xRSH + R′ Sx RSm−x MPC where x and m are the numbers of new and original ligands [188], respectively. The rate and equilibrium stoichiometry, x, of this reaction, are controlled by factors such as the mole ratio of R′ SH to RS, their relative steric bulk, and R versus R′ chain lengths [199]. The stepwise place-exchange reaction additionally reveals the extent to which an exchange reaction can displace —functionalized versus nonfunctionalized alkanethiolate ligands from a cluster. An important observation is that ligands with longer chains tend to displace those with short linker chains, but the displacement of ligands with long linker chains by those with short linker chains is less likely to occur. Also, short chain length, bulky, -functionalized alkanethiolates are the least thermodynamically stable ligands, consistent with the fact that they have a varying density of chain packing around the gold core [199]. Detailed studies of the dynamics and the mechanism revealed that the place-exchange reaction (a) has a 1:1 stoichiometry, (ii) is an associative reaction, (iii) yields the displaced ligand in solution as a thiol, and (iv) does not involve disulfides or oxidized sulfur species [198]. The timedependent rate of place exchange has been interpreted as reflecting a hierarchy of different core surface binding sites with associated susceptibility (vertexes, edges, terraces) to place exchange [193, 198, 199]. These observations are similar to the displacement of alkanethiols on self-assembled monolayers on flat substrates [204].

36 3.3. Chemisorption of Alkanethiols As mentioned, surface modification of gold nanoparticles by chemisorption of alkane thiols is a logical choice for surface functionalization of gold nanoparticles provided that the proper steps to avoid irreversible aggregation are taken. The interaction of alkanethiols with the gold nanoparticle surface causes the thiol molecules self-assemble in a well ordered, densely packed manner on the nanoparticle surface (similar to SAMs on flat substrates). Common precursors of these SAMs are aliphatic thiols with the structure of HS–(CH2 n R, where the functional group, R, stands for a variety of functionalities [172–181]. In principle, these groups can allow for further functionalization of the gold nanoparticles with ligands, proteins, and other biomolecules. Direct chemisorption of alkanethiols on already synthesized colloidal gold produces functionalized nanoparticles similar to MPCs. However, there are some disadvantages associated with this method. Under some conditions (pH, ionic strength, alkane thiol chain length, and alkane thiol functionality) and during their functionalization, the nanoparticles can undergo irreversible aggregation or flocculation [172, 173]. This limits the use of this method for surface modification purposes. Although used to some extent by several research groups [25, 205–207], the conditions under which this method is employed usually require dilute solutions and the modified particles lack stability over long periods of time or under extreme conditions of pH, ionic strength, and temperature. Our group has recently proposed a method that greatly facilitates the formation of SAMs of alkanethiols on colloidal gold [208]. The approach to obtain stable alkanethiol modified gold nanoparticles using this method consists of performing the chemisorption of carboxyl terminated alkane thiols in the presence of a nonionic surfactant [polyoxyethylene (20) sorbitan monolaurate (Tween 20)]. This method was developed from the observation that the physisorption of Tween 20 prior to chemisorption of alkanethiols provides a protective, stabilizing layer around the nanoparticles preventing them from aggregation by means of steric hindrance caused by the oligo(ethylene glycol) moieties on this surfactant molecule. Since the physisorption of Tween 20 on the gold nanoparticles is weak compared to the chemisorption alkane thiols, the weakly adsorbed surfactant can be displaced by alkanethiols chemisorbing onto the gold surface. The alkane thiol modified gold nanoparticles obtained by this method have remarkable stability in the sense that they can be frozen or dried and resuspended with mild sonication without suffering irreversible aggregation. Further derivatization of these particles with other molecules is possible and it is accomplished easily without formation of particle aggregates. Often, surface functionalization with alkanethiols is only an intermediate step toward further functionalization of nanoparticles. This is the case for the place exchange reaction and further derivatization of chemically reactive groups. [182, 198–203] Another strategy involves a noncovalent coupling of ligands by means of hydrophobic interactions that consists on the formation of interdigitated bilayers [209–212]. The latter appears to be versatile enough for derivatization with a variety of compounds and the

Colloidal Gold

modified surfaces are remarkably stable in aqueous environments which makes this method attractive for biological applications.

3.4. Proteins and Antibodies (Immunogold) Identification and localization of components of biological systems (e.g., visualization of proteins within a cell) using gold nanoparticles attached to an antibody that binds specifically to a protein is a commonly used method in electron microscopy studies of cells. The combination of the reliable immunological methods for producing good quality monoclonal and polyclonal antibodies and the techniques for complexing proteins onto gold nanoparticles to form immunological probes allows the investigation of a vast range of antigens in cells and tissues. These gold conjugates can be used in a wide range of applications. The major reason for their use in electron microscopy studies is the high electron density of the gold nanoparticles coupled with the ease with which different particle sizes can be used for examination at different magnifications. Also the strong emission of secondary electrons and backscattered electrons from gold nanoparticles make the gold probes ideal for study of surface antigens and macromolecules with scanning electron microscopy (SEM) [213]. Immunogold probes give the highest possible spatial resolution to obtain structural information at the macromolecular level. Also nanogold is known to produce a more punctuate labeling pattern than eosin photo-oxidation and thus has advantages in electron microscopic quantitation and resolution of binding sites. Immunogold gives more intense stains compared to other stains and this permits much higher dilutions of primary antibody and gold conjugate, resulting in lower background. Colloidal gold labeling techniques were first introduced by Faulk and Taylor when they absorbed antisalmonella rabbit gamma globulins to gold particles for one-step identification and localization of salmonella antigens [214]. Indirect labeling techniques with gold probes were subsequently introduced by Romano et al. who also reported gold labeling with protein A for detection of primary immunoglobulins [215]. The application to thin sections for electron microscopy was described in detail by Roth et al. and since then colloidal gold has been used in TEM for ultrastructural studies of cellular antigens [216]. Electron imaging of gold particles has also been combined with X-ray emission using an energy dispersive analyzer in SEM for immunolabeling studies [217]. To date, the method of application of gold probes to the tissues has been well documented [218]. Gold probes in sizes ranging from 1 to 40 nm have been in use for electron microscopy. But the possibility of ambiguity between small gold particles and tissue structures indicates that larger particle sizes are preferred and are best visualized by backscattered electron imaging. While all sizes of gold probes may be used to label tissue proteins, the sizes most commonly employed for SEM studies are 20–30 nm. Gold nanoparticles may be conjugated to primary antibodies for one-step identification of antigens (e.g., with anti Human IgG) but are more usually employed as secondary antibody/protein labels. Also multiple labeling of different

Colloidal Gold

cellular proteins on the same structures using different sizes of gold probes or gold probes in conjunction with other labels has become a popular method [219, 220]. On cells of eukaryotic origin, which have a net negative surface charge from anionic plasma membrane components, the membrane charge distribution is thought to be important in the movement of various soluble macromolecules across cell walls. Various methods have been employed to analyze the spatial distribution of anionic sites in tissues and to study the role played by surface charges in intracellular and intercellular dynamics. Cationic colloidal gold is one of the most used probes for this study [221–224]. Cationic gold results from the conjugation of colloidal gold particles with poly-l-lysine, which exhibits a strong positive charge at physiological pH [223]. A one-step incubation with these conjugates reveals subcellular sites bearing a net negative charge and such gold probes may be used at physiological pH and ionic strength. Cationic gold can also be tested for use in paraffin and resin embedded sections. Gold cluster complexes offer several advantages over conventional colloidal gold for labeling. A number of different reactive groups may be incorporated during synthesis and used for covalent, site-specific attachment to biomolecules. An example of this nanogold cluster is a gold core of 1.4 nm in diameter, stabilized by tris(aryl)phosphine ligands. This is a neutral molecule, stable over a wide range of pH and ionic concentrations. It has been used to label IgG molecules, Fab’ fragments, streptavidin, and other proteins, as well as smaller molecules such as peptides [225], oligonucleotides [226], and lipids [227], which cannot be directly labeled with colloidal gold.

3.5. Dendrimers and Polymers Dendrimers and polymers can also be used to functionalize gold nanoparticles. Section 2.2 provided a description on the use of polymers for the preparation of colloidal gold. From that, it is obvious that gold nanoparticles can also be functionalized with polymers and examples of this abound in the literature. Citrate-capped gold nanoparticles have been efficiently grafted with nonionic hydrophilic polymers bearing hydroxyl groups, using disulfide anchors by the “grafting-to” strategy (in which a covalently attached polymer monolayer is formed by simple contact of the metal surface with dilute solutions of polymers) enabling their use in biological applications [228]. The resulting monolayers can be derivatized for specific recognition of analytes using unsymmetrical bifunctional linker groups or directly when the polymer bears activated sites. Two monomers used in this approach are N -[tris(hydroxymethyl)methyl]acrylamide (NTHMMAAM) and N -(isopropyl)acrylamide (NIPAAM). Derivatization of the gold surface by the disulfide bearing polymers (NTHMMAAM or NIPAAM) improved the stability of nanoparticles toward aggregation. They remained water-soluble during all the synthesis steps and could also be lyophilized and redispersed in water, indicating the complete protection of gold colloids by the polymer monolayer. Additionally, gold nanoparticles coated with the thermoresponsive polymer poly(NIPAAM) underwent conformational collapse with increased temperature, resulting

37 in controlled and reversible thermally induced aggregation [228]. End functionalized 3D SAMs on gold nanoparticles have been used to initiate a living cationic ring-opening polymerization reaction directly on gold nanoparticle surfaces, by the “grafting-from” technique (in which a reactive group is created on the surface that is able to initiate the polymerization, and the propagating polymer chain is growing from the surface) [229, 230]. Using appropriate initiators and 2-oxazoline monomers (2-phenyl-2-oxazoline/2-ethyl2-oxazoline), ring-opening polymerization can proceed in a living manner in a “one-pot multistage” reaction, creating complex core–shell morphologies of amphiphilic nanocomposites [231]. This combination of a grafting– from reaction resulting in polymer “brush-type” shells of linear macromolecules and the introduction of a terminal mesogen by means of a termination reaction can result in stable, amphiphilic, core–shell materials with a well defined hydrophilic/lipophilic balance. Based on the synthetic concept developed for preparing dense polymer brushes on planar gold substrates [232, 233], the exposure of 11-hydroxyundecane-1-thiol (HUT)-functionalized gold nanoparticles to a trifluoromethanesulfonic anhydride vapor/nitrogen stream results in the conversion of the HUT hydroxyl groups to triflate functional groups. Addition of dry chloroform and the freshly distilled monomer (2-ethyl-2-oxazoline or 2-phenyl-2-oxazoline) at 0  C under vigorous stirring followed by subsequent heating in an oil bath allows for polymerization to take place on the surface of the nanoparticles. The polymerization can be terminated by addition of N , N -di-n-octadecylamine (for 2-ethyl-2-oxazoline) solution in dry chloroform or piperidine [for poly(2-ethyl-2-oxazoline)] at 0  C. The surface-bound triflate SAM is a highly efficient initiator for the cationic ring-opening polymerization of oxazolines and the triflate gegenion ensures a strictly ionic mechanism of the propagation reaction, ensuring a living and stoichiometric polymerization [232, 233]. A similar procedure for the preparation of terminally functionalized poly (N -propionylethylenimine) (PPEI) by cationic living ring-opening polymerization of 2-ethyl2-oxazoline in a one-pot multistep reaction, by the so called “grafting-onto” approach (which involves the chemisorption of macromolecules), involves a direct combination of the fast initiation technique and a selective quantitative termination reaction [234]. The chain ends of the linear hydrophilic polymers can be further functionalized with different lipophilic moieties (methyl-, n-hexadecyl- (C16 , and 1,2-O-diooctadecyl-sn-glyceryl- [(C18 )2− ]) and a monofunctional silane-coupling group. By varying the length of the double functionalized hydrophilic PPEI chain n = 10, 20), surface active amphiphilic lipopolymers of controlled hydrophilic–lipophilic balance for surface functionalization can be obtained in high overall yields [234]. The fabrication of surfaces with controlled modulation of interfacial properties (e.g., wettability) involving the adsorption or grafting of “smart” (environmentally responsive) polymers onto surfaces has also been implemented on metallic nanoparticles. Chilkoti et al. [235]. used the change in optical properties of colloidal gold upon aggregation to develop an experimentally convenient colorimetric method

38 to study the interfacial phase transition of an elastinlike polypeptide (ELP). This thermally responsive, genetically encoded biopolymer was covalently attached to mercaptoundecanoic acid functionalized gold nanoparticles. Raising the solution temperature from 10 to 40  C thermally triggered the hydrophilic-to-hydrophobic phase transition of the adsorbed ELP resulting in formation of large aggregates due to interparticle hydrophobic interactions, causing a color change from red to violet. Below the phase transition temperature, the Au–ELP colloids in suspension have an average interparticle distance several times greater than the colloid radius and hence appear red. Upon raising the temperature of an Au–ELP suspension, the adsorbed ELP undergoes an intramolecular hydrophilic to hydrophobic transition and the interparticle distance of the colloids in the aggregated state approaches the colloidal radius, which results in a redshift in the absorbance spectrum (vide supra, Section 1.2), and the suspension consequently appears violet. Cooling the suspension back to 10  C causes the ELP to transition back to the hydrophilic state, which dissociates the colloidal aggregates and returns the absorbance maximum to 535 nm. Figure 5 shows the schematic of the reversible aggregation of ELP-functionalized gold nanoparticles [235]. In addition to linear polymers, hyperbranched molecules called dendrimers can also be used to functionalize gold nanoparticles. Dendrimers are attracting increasing attention because of their unique structures and properties [236–243]. These aesthetically appealing synthetic macromolecules are monodispersed and hyperbranched polymers constructed from ABn monomers (AB is a monomer, n usually 2 or 3) rather than the standard AB monomers, which produce linear polymers. Their synthesis in an iterative fashion leads to a nonlinear, stepwise synthetic growth wherein the number of monomer units incorporated in each successive iteration roughly doubles (AB2  or triples (AB3  that in the previous cycle. Thus each repetition cycle leads to the addition of one more layer of branches—called a generation—to the dendrimer framework. The generation

Figure 5. Reversible aggregation of gold nanoparticles caused by the environmentally triggered phase transition of adsorbed ELP: (A) Au– ELP in the hydrophilic phase with spectra typical of isolated gold colloid and (B) aggregates which exhibit a redshift in their absorbance spectra compared to that of individual colloid in (A). Reprinted with permission from [235], N. Nath et al., J. Am. Chem. Soc. 123, 8197 (2001). © 2001. American Chemical Society.

Colloidal Gold

number of the dendrimer is equal to the number of repetition cycles performed and may be easily determined by counting the number of branch points as one proceeds from the core to the periphery [236–243]. Dendrimer stabilized nanocomposite materials have unique potential for applications such as catalysis which require stable colloidal particles having surfaces that are not completely passivated by a stabilizing adsorbate. Because of the low mass density within the dendrimer interior, such monolayers are highly permeable to small molecules [244]. Thus, dendrimers can serve the dual purpose of acting as a “nanofilter” that passes small and specifically charged species, prevents other molecules from interacting with the colloid, and can function as a stabilizer. Dendrimers are also recognized as monodispersed nanoreactors, possessing architectures that allow the preorganization of metal ions within their interiors [245]. Several noble metal nanoparticles have been prepared in the presence of dendrimers, in which the size of the nanoparticle can be controlled by the size of the dendrimers [246–250]. Two molecules of this type that are frequently used are polyamidoamine (PAMAM) dendrimers and polypropylene imine dendrimers. Au colloids in the 2–3 nm size regime have been prepared by one-phase in-situ reduction of an aqueous solution of HAuCl4 in the presence of either second- or fourth-generation (G2 or G4, respectively) PAMAM dendrimers using NaBH4 as the reducing agent. The dendrimers encapsulated the colloids, imparting stability to the aqueous colloidal solutions. The driving force for the interaction of the colloids with the dendrimers was an association of Au with the primary amine terminal groups (and perhaps the interior secondary and tertiary amines) of the dendrimer. The dendrimer generation used in such syntheses controlled the size of the resultant colloids: lower generation dendrimers gave rise to larger, less monodisperse and more aggregated colloids than those prepared using higher generation dendrimers, for a constant ratio of primary amine groups to AuCl4− . Also the stability of nanoparticles decreased sharply when either the solvent or the excess free dendrimer was removed from the solution [247, 251]. Sugar-persubstituted PAMAM dendrimers (sugar balls) were synthesized by Aoi et al. [252]. Since sugar balls have highly ordered structure with arranged saccharides on the peripheries of the dendrimers and terminal hydroxyl residues that can operate as reductants, sugar balls were used to act as protective agents for gold nanoparticles, against flocculation. When Au3+ ions are reduced with the hydroxyl groups of the sugar balls, the hydroxyl groups are oxidized to carbonyl groups. Thus it is found that sugar balls can act as protective agents as well as reductants for the preparation of gold nanoparticles [253]. Fourth-generation PAMAM dendrimers [254] having terminal groups partially or fully functionalized with thiol groups have also been explored for the functionalization and synthesis of gold nanoparticle [255]. Diluted solutions of thiolated dendrimers and tetrachloroauric acid can be mixed and reduced with NaBH4 to produce stable dendrimer functionalized gold nanoparticles. The enhanced stability with thiolated dendrimers compared to amine-stabilized materials is due to the low concentration of dendrimer required to prepare the nanocrystals.

39

Colloidal Gold

3.6. Sol–Gels

4. CHARACTERIZATION

As explained in the previous sections, surface modification of gold nanoparticles imparts these systems with a variety of functionalities that makes them useful on a wide variety of applications. Surface modification can be achieved as a series of sequential steps starting with the synthesis of the particles or in a one-step procedure. For most applications the ligands are selected such that they are used both as stabilizing agents and as functional surface groups for the immobilization of other molecules onto the nanoparticle surface. With the availability of a variety of ligands, the choice of ligands is almost entirely dependent on the application in which the gold nanoparticle will be used. Most of the time, alkane thiols, polyelectrolytes, and dendrimers are used for the purpose of both surface modification and stabilization of gold nanoparticles. Often, the ligands employed may affect the electronic (optical, chemical) properties of the nanoparticles, for example, adsorption of a nucleophilic molecule on gold nanoparticles results in a redshift in surface plasmon resonance and an increase in absorbance at higher wavelengths [22, 99]. Alternative methods of preparing surface modified gold nanoparticles have been explored such as functionalization with silica by sol–gel processing. Silica has recently been utilized for the stabilization of metallic nanoparticles [256–261]. Sol–gel processing is a powerful technique that offers unique opportunities for the synthesis of optical materials in the form of thin films, fibers, and fine powders [262]. Several general approaches have been reported for the functionalization of gold nanoparticles with silica by sol–gel processing. Martino et al. used sol–gel processing in reverse micelle solutions and sequential reduction of gold salt for the creation of gold–silicate nanodispersions [263]. However, the presence of gel precursors destabilizes the inverse micelles resulting in larger nanoparticles and wider particle size distributions. Liz-Marzan et al. synthesized silica coated gold nanoparticles in a two-step process, where gold nanoparticles are first reduced by citrate and the nanoparticles are coated with silica using a silane coupling agent [256]. During the synthesis of silica coated gold nanoparticles a monolayer of a silane coupling agent [e.g., (3-aminopropyl) trimethoxysilane or N -(3-trimethoxysilyl) propyl ethylenediamine] is allowed to adsorb onto gold nanoparticles where the strong affinity of gold for amine groups is exploited [187, 256, 261]. In a second step, active silica (sodium silicate) is added to the dispersion promoting the formation of a thin, dense, and relatively homogeneous silica layer around the nanoparticles using the silanol groups as anchor points. At this stage, the nanoparticles can be transferred into ethanol and the silica layer thickness can be increased in a controlled way. With this method, the morphology and the concentration of gold nanoparticles over a wide range can be controlled without sacrificing the stability, and the optical properties of the gold nanoparticles agree extremely well with the predictions of Mie theory [256]. Thus, the preparation and surface modification of gold nanoparticles via sol– gel processing is another alternative to the well-established gold–alkane thiol method. Moreover, similar procedures can be used for other nanoparticles such as platinum, silver, and palladium and semiconductor particles.

A variety of techniques are used to characterize gold nanoparticles. In general a combination of techniques is needed in order to characterize these systems because no single analytical technique can provide complete information on nanoparticles. One of the most commonly used techniques in the study of colloidal gold is UV-vis spectroscopy. The absorption maximum, which is due to excitation of surface plasmons on the metallic nanoparticles, is observed around 520 nm for spherical gold nanoparticles. The position of this peak is only weakly dependent on the size of the gold nanoparticles [22]. However, changes in the dielectric medium around the particles can have a significant effect on the position of the SPR peak, shifting to longer wavelengths with increased values of the dielectric constant of the medium. Additionally, the SPR position also depends on the shape of the particles and their proximity to each other [17, 18, 22–25, 139]. When the separation distance between metallic nanoparticles is smaller than their radius additional resonances occur at longer wavelengths. Thus, if the individual particles approach each other to within less than one particle diameter the absorption spectra broadens and shifts to longer wavelengths. These changes in the absorption spectrum are routinely used to determine the presence of interparticle interactions such as aggregation and have also been exploited for DNA detection. Optical absorption spectroscopy can give semiquantitative information on colloidal gold aggregation. However, the state of aggregation and the size distribution of the nanoparticles can be more accurately assessed with TEM. TEM is a very effective (and commonly used) technique for the characterization metallic nanoparticles due to their high density of free electrons. TEM gives immediate visualization of the gold nanoparticles by capturing the contrast created when a high-energy electron beam is passed through them. As electrons hit the nanoparticles on the sample grid, some of them will be scattered while the remainder are focused onto a photographic film to form an image. Unfocussed electrons are blocked out by the objective aperture resulting in an enhancement of the image contrast. The resolution limits of modern TEM are sufficient enough to image gold nanoparticles in the 1–10 nm size range. Although atomic resolution and lattice imaging are required for detailed structural analysis, a particle size distribution can be obtained at moderately low resolution for gold nanoparticles with mean diameter less than 10 nm [264–268]. Since modern applications of colloidal gold have exploited the capability to functionalize their surfaces with a variety of ligands the composition of the nanoparticles also needs to be characterized. It may be necessary to find the oxidation state of gold or the entire composition of the monolayer. X-ray photoelectron spectroscopy (XPS) is a surface sensitive technique that can provide this information [269]. Additionally, XPS can help determine the presence of contaminants remaining from the preparation procedures for gold nanoparticles (e.g., adsorption of unreduced gold cations and their corresponding anions during gold salt reduction). XPS can also be used to determine the composition of these species with high accuracy [268, 270].

40 Nuclear magnetic resonance (NMR) can also be used for the characterization of gold nanoparticles. This technique is based on the difference between the resonant frequency of a metal and a diamagnetic compound such as a complex or ionic solid. This difference (NMR shift or the Knight shift) is a consequence of the interaction of a metal nucleus with the conduction electrons and is measured as a fraction of the applied field. The NMR shift for [197] Au is +14%. This means that in a field in which a standard such as [PtI6 ]2 resonates at 100 MHz, the 197 Au resonance for metallic gold is shifted by 1.4 MHz. The measurement of the NMR shift of gold in small particulate form provides crucial information on the size at which gold nanoparticles begin to develop metallic properties, which is an important topic in nanotechnology. Moreover, NMR spectroscopy is particularly informative about the structure and chemical composition of MPCs and alkane thiol modified gold nanoparticles [180, 182, 183, 186, 271, 272]. For monolayer characterization 1 H and 13 C NMR resonances are usually measured, since mostly hydrogen and carbon are present in the monolayers adsorbed on the gold clusters/nanoparticles. In this case, strong NMR shifts are observed due to the interaction of the adsorbate nuclei with the conduction electron density of the gold surface. Fourier transform infrared spectroscopy (FTIR) can also be used to elucidate the surface chemistry of molecules adsorbed on gold nanoparticles [272–274]. Information on the order–disorder (gauche defects) of the chains of the molecules or the defects near gold nanoparticle surface is obtained by measuring the vibrational energy levels of each bond (C–H, C–C, C–N) in the infrared region of the spectrum. These energy modes are characteristic to each bond and shifts in the vibration modes provide the necessary information on the structure and content of the molecules adsorbed on gold nanoparticles. Fourier transform infrared spectroscopy can provide valuable information on the conformation and order of alkyl chains on monolayers, particularly through analysis of the (CH2  bands. However, symmetrical bonds such as C–C appear as very weak bands or may not be detectable at all by FTIR. Thus, FTIR normally does not produce this information [275]. Symmetrical bonds, however, are readily detected using Raman spectroscopy. Thus, this technique can provide complementary information on the C–C backbones of monolayers. Other bands present in alkanethiol SAMs such as C–S and S–H bands can be analyzed easily by Raman spectroscopy, thus providing structural and chemical insight into the interaction between the alkanethiolates and the metal substrate [128, 246, 275–277]. For example, the S–H bond of alkanethiols is cleaved upon chemisorption to gold [275]. Thus, the relative intensity of the (S–H) band of free thiol and gold thiolates samples can indicate the progression of alkanethiol chemisorption. The (C–S) band in the Raman spectrum also indicates the chemisorption process and the resultant structure of the alkyl chains near the surface of gold [275, 278]. Thus, FTIR and Raman spectroscopy provide complementary information on the vibration modes in the chemisorbed monolayers on metallic nanoparticles. In this context, it is a fortunate fact that metallic nanoparticles further enhance the Raman signals [128, 146].

Colloidal Gold

5. APPLICATIONS Although colloidal gold has been used for hundreds of years, the number of applications has been limited until recently. For centuries, its main applications were as a stain for glass and fabrics and to a minor extent as a purported therapeutic agent. However, there is currently a boom on its applications, particularly within the realm of nanoscale science and engineering. In the following, a number of newer applications developed within the last decade are described. It is likely that newer applications will be developed because of the large interest in nanotechnology.

5.1. Colorimetric Assays One of the most remarkable applications of gold nanoparticles reported in recent years is the colorimetric detection of oligonucleotides based on the optical properties of the gold nanoparticles [206, 279]. This simple and highly selective detection method has many advantages over other sequencespecific DNA detection methods currently applied in the diagnosis of pathogenic and genetic diseases. Most of the current systems make use of the hybridization of an immobilized target oligo- or polynucleotide probes labeled with radioactive 32 P or 35 S reporter groups [280–283]. Although this technique offers high sensitivity, the use of radioactive labels creates certain problems such as the disposal of the probes, the requirement of specially trained personnel, and the short shelf life. In other techniques, the radioactive labels are replaced with nonradioactive labels such as organic dyes, and the hybridization of target DNA molecules is detected by fluorescence or luminescence spectroscopy with high sensitivity [280–283]. However, the labels and the monitoring equipment are relatively expensive increasing the cost of the method. The detection of oligonucleotides using gold nanoparticles has many desirable features such as rapid detection, a colorimetric response, and little or no required instrumentation [206, 284]. This approach involves two gold nanoparticle probes with covalently bound oligonucleotides that are complementary to a target of interest. When exposed to target strands, these nanoparticle probes are cross-linked through DNA hybridization and form networks of aggregates composed of thousands of nanoparticles. The aggregation process is accompanied by red-to-blue color change of the solution, which occurs due to changes in the SPR frequency of the gold nanoparticles. The detection mechanism is based upon the fact that when the interparticle distance between the nanoparticles is greater than the average nanoparticle diameter, the color of the aggregated appears red, but as the interparticle distance decreases the color shifts to blue providing a means of detection. Also, the melting characteristics of these networks (they have sharp melting transitions) allow one to differentiate a perfectly complementary target strand from a strand with a single base mismatch. Moreover, the color changes associated with hybridization are significantly enhanced and easily visualized even with the naked eye if the hybridized sample is processed on a solid support [284]. This method provides a simple, fast, qualitative response for the detection of mismatches in the DNA sequence [206, 284–286].

41

Colloidal Gold

With this approach, quantitative responses can be obtained using larger gold nanoparticles [285] and incorporating silver nanoparticles into the system that allows the responses to be obtained by an inexpensive flatbed scanner [286, 287]. Since the molar extinction coefficient of 50 nm gold nanoparticles is 1011 M−1 cm−1 (108 M−1 cm−1 in case of 13 nm nanoparticles), the use of larger gold nanoparticles provides higher sensitivity assays and allows quantitative colorimetric detection. It was shown that increasing the nanoparticle size resulted in significant improvement in the extinction dampening of the surface plasmon band (indicative of oligonucleotide hybridization-induced nanoparticle aggregation) as evidenced via melting analyses. A detection range of 50 pM to 5 nM was obtained [286]. Another way to increase the sensitivity of detection and quantify the hybridization process is by incorporation of silver (Ag) to gold (Au) nanoparticles in a so-called “core–shell” scheme. In this method, silver ions are reduced to silver metal on gold nanoparticles creating core–shell Au/Ag nanoparticles that can be easily modified alkylthiol-oligonucleotides. The molar extinction coefficient of the core–shell nanoparticles is very similar to that for the silver nanoparticles alone and is larger than that for gold nanoparticles of the same size. These oligonucleotide-modified core–shell nanoparticles show similar but enhanced changes in the optical spectrum and has similar melting properties as oligonucleotide-modified gold nanoparticles when they are aggregated due to hybridization. This process increases the scanned intensity by a factor as large as 105 and target oligonucleotides concentrations as low as 50 fM can be measured [287]. The strong dependence of the SPR absorption band on the degree of aggregation of gold nanoparticles has also been exploited to detect heavy metals. Various metal ions like zinc, manganese, nickel, lead, cadmium, and mercury may be present in water at parts per million concentrations or higher. Some of these are toxic metal ions that pose significant public health hazards when present in drinking water. For absolute identification and total concentration assessment, ion-coupled-plasma spectroscopy is commonly used as a characterization method. However, more efficient detection can be implemented using colloidal gold. In gold nanoparticles covered with capping agents like mercaptocarboxylic acid, the surface carboxyl groups can enable chelation of metal ions [288–290]. The principle of metal-cation based aggregation has been used by Murray and co-workers to initialize the formation of metal films and solution-phase aggregates [291, 292] similar to the aggregation of DNA-functionalized metal nanoparticles that has been used for the sequence-specific detection of the oligonucleotides [206, 279]. A colorimetric technique, particularly appropriate for otherwise largely spectroscopically silent species like trace mercury, cadmium, and lead has been demonstrated using the principle of change in color from red to blue due to the ionchelation based aggregation of functionalized gold nanoparticles. The aggregation process produces a shift in plasmon band energy and a substantial increase in long wavelength Rayleigh scattering, which has been employed for visual detection of these ions. In this method functionalized gold nanoparticles are aggregated in solution in the presence of divalent metal ions by an ion-templated chelation process.

This caused a measurable change in the absorption spectrum of the particles. The aggregation also enhanced the hyperRayleigh scattering (HRS) response from the nanoparticles solutions, providing an inherently more sensitive method of detection. The increase in the HRS intensity appeared to be due to both symmetry lowering and an increase in the effective size of the chromophoric entity responsible for plasmonenhanced HRS [293]. One of the challenges in devising colorimetric methods for the identification of these ions has been the identification of appropriately strongly binding leuco dyes capable of yielding sufficiently intensely absorbing metal/dye complexes. Thus appropriately functionalized gold nanoparticles could be used as exceptionally high extinction dyes for colorimetric sensing of heavy metal ions [293]. Rao et al. [294] investigated the effect of concentration of metal ions on the interaction of nanoparticles and chelation of ions. Figure 6 shows the ion-chelation mechanism binding two nanoparticles. Gold nanoparticles capped with -lipoic acid were interacted with various metal ions like Cu, Fe, Ni, Mn, Cd, Zn, and Pb. It was found that in the dilute metal-ion concentration regime the metal ions bound to the carboxyl functional group of the -lipoic acid and thereby caused observed dampening of the plasmon band. The chelation/aggregation process was reversible via addition of a strong metal ion chelator such as EDTA. The intermediate metal-ion concentration regime was characterized by the aggregation and irreversible chelation accompanied by an increase flocculation. Further increase in concentration resulted in precipitation. Another study by Szulczewski et al. [295] on the adsorption of Hg atoms on the gold and silver nanoparticles revealed the influence of particle size on the properties of the nanoparticles. The adsorption of Hg on nanoparticles resulted in a blueshift of the surface plasmon mode, the shift being more pronounced for the smaller particles than the larger particles for the same Hg shell thickness. More research in this field is to be done in terms of improving the modest colorimetric detection limits, enhancing the chemical selectivity and sensitivity, and study of properties in the presence of metal ions.

5.2. Fluorescence Assays Besides their unique optical properties, metallic nanoparticles metals can also exert an important influence in the fluorescence emission characteristics of fluorophores positioned in their vicinity. Two different phenomena can be observed depending on the separation distance between the surface of gold nanoparticles and the fluorophores. In general, when the separation distance is smaller than 5 nm,

Figure 6. Ion–chelation binding of two nanoparticles. Reprinted with permission from [294], S. Berchmans et al., J. Phys. Chem. B 106, 4647 (2002). © 2002, American Chemical Society.

42

F F

F

Emission Intensity (Normalized)

F

strong quenching of the emission of fluorescence and dramatic reduction in the lifetimes of the excited states of fluorophores is observed [296–299]. On the other hand, when the separation distance is between 10 and 20 nm, enhanced emission of fluorescence is observed due to local concentration of the incident excitation field by the gold nanoparticles [299–303]. Although these phenomena are well known, to date, only a very limited number of applications involving nanoparticles are reported in the literature for the detection of molecules [303, 304]. In 2002, Maxwell et al. [303] reported the use of gold nanoparticles in the detection of DNA sequences and singlebase mutations based on the quenching of fluorescence emission of fluorophores that are attached to the oligonucleotides. The gold nanoparticles were 2.5 nm in diameter and functioned as a “nanoscaffold” for the attachment of fluorophore-labeled oligonucleotides and as a quencher (or energy acceptor). The fluorophore-labeled oligonucleotides assembled into a constrained archlike conformation on the gold nanoparticle surface such that the thiol group and the fluorophore adsorb on the same nanoparticle. In this state, the fluorescence emission of the fluorophore is quenched by the gold nanoparticle. Upon binding of the target DNA, the constrained conformation opens, and the fluorophores separate from the surface because of the structural rigidity of the hybridized DNA, leading to an increased emission of fluorescence. Desorption of the fluorophore from the gold surface was attributed to the fact that the hybridization energy of the DNA ($H = 80–100 kcal/mol) is considerably larger than that of adsorption energies of organic dyes on gold ($H = 8–16 kcal/mol) [305]. Since the separation distance between the fluorophore and the gold nanoparticle in the open state is about 10 nm, nonradiative energy transfer from the excited fluorophores to gold nanoparticles was considerably decreased upon hybridization because when the nanoparticles are small (2–3 nm), nonradiative energy transfer occurs within 1–2 nm from the gold nanoparticle surface, thus resulting in a dramatic increase of fluorescence upon hybridization [306, 307]. Our group has researched how to use gold nanoparticles as platforms for displaying ligands in fluorescence-based assays that exploit the quenching phenomena exerted by gold nanoparticles on fluorophores. For this, alkanethiol functionalized gold nanoparticles are used in this detection scheme because the thickness of the SAMs is in the order of 1–3 nm [174–177], which positions the fluorophores in a region from the gold surface where the fluorescence emission is more sensitive to changes in the separation distance [296–299]. Proof on concept demonstration of this biosensing system is shown in Figure 7. The specific interaction of biotinylated gold nanoparticles with fluorophore-labeled antibiotin caused a decrease in the fluorescence emission intensity. Upon addition of soluble d-biotin, competitive dissociation of bound antibiotin caused a recovery in the fluorescence emission intensity. When the binding sites of antibiotin were saturated with d-biotin prior to the interaction with biotinylated gold nanoparticles, there was no change in the fluorescence emission, which further shows the specific nature of the interactions taking place between the biotinylated gold nanoparticles and the labeled antibody.

Colloidal Gold

1.0 blocked anti-biotin 0.9

0.8

non-blocked antibiotin

0.7

0.6 06

00

1200

1800

2400

Time (seconds) Figure 7. Binding and dissociation of fluorescently labeled antibiotin to biotinylated gold nanoparticles.

The biosensing scheme described could be easily generalized to other ligand–protein pairs where specific molecular interactions at the surface of the gold nanoparticles cause the separation of fluorophore labeled biomolecules from the gold nanoparticle.

5.3. Monolayers In addition to their use in surface-enhanced Raman scattering gold nanoparticles can have uses in electro-optics and molecular electronics in the form of thin films. These applications require the assembly of the nanoparticles on solid surfaces in the form of desired one-, two-, and three-dimensional structures. Some of the strategies for the formation of gold nanoparticle thin films on solid surfaces take advantage of the ionic nature of gold [308] as well as its affinity toward thiol or amine molecules, while in some applications polymers such as poly(allylaminehydrochloride), poly(sodium 4styrene sulfonate), and poly(ethylene imine) are used [309]. Natan and his research group exploited one of the earliest approaches to assemble gold nanoparticles onto chemically functionalized substrates [264–266, 308, 310]. In this approach, organosilanes with different functional groups are used as the linker between the glass substrates and the gold nanoparticles. Gold nanoparticles are strongly bound to the surface through covalent bonds to pendant functional groups such as CN, NH2 , or SH. The general assembly protocol is carried out completely in solution: clean glass substrates are immersed in methanolic solutions (or in toluene) of organosilanes, rinsed, and subsequently immersed in aqueous gold nanoparticle solutions. Two-dimensional arrays spontaneously form on the organosilane surface. The resulting substrates can be characterized by UV-vis spectroscopy, TEM or atomic force microscopy, and SERS. The flexibility and control that this method provides are noteworthy, such that by manipulating the particle size, particle–silane interactions and the physical and chemical properties of the substrate the nanoscale architecture can be controlled. This solution-based assembly protocol makes the substrate

43

Colloidal Gold

fabrication routine and it also allows one to fabricate substrates with any shape or size. Therefore, gold nanoparticles monolayers can be prepared on glass/quartz slides and TEM grids. The same strategy has also been used by other groups [311–313]. Another approach to making thin films of gold nanoparticles on solid substrates is by functionalization of both substrate and the nanoparticles and chemically coupling them via functional groups [273]. Gold nanoparticles can be functionalized with a ligation partner such as ketone and then chemoselectively immobilizing them onto planar gold substrates presenting SAMs of aminooxy groups. Ketone and aminooxy groups react selectively to form a chemically stable oxime, and they are unreactive toward thiol that is essential for the preparation of gold nanoparticles as well as the monolayer film. The resulting films of ketone-functionalized gold nanoparticles are fairly stable and can be characterized with ellipsometry, grazing-angle infrared spectroscopy, and electrochemistry. Since this technique requires mutually exclusive reactivity of pairs of functional groups, even among a number of potentially reactive functional groups, only the two chemoselective ligation partners will react with each other. Because the reaction can tolerate the presence of a diverse functionality so that protecting-group manipulations are unnecessary, it has been also used to conjugate peptides and proteins onto surfaces [314]. Metal nanoparticle loading on surfaces can be increased by the use of additional linker molecules to bind more nanoparticles to the surface [265, 292, 315] by the modification of the surface by preadsorption of polyelectrolytes (layer-by-layer adsorption) [316, 317] or by repeated washing and reexposure of the substrates to nanoparticles solutions [318, 319]. In the layer-by-layer adsorption method, functionalized molecular linkers are first immobilized onto a substrate followed by dipping into a solution of unmodified gold nanoparticles to immobilize a layer of nanoparticles on the linker layer. The resulting nanoparticle layer is then immersed again into the solution of linkers for further immobilization. Repeating sequentially the adsorption processes results in a multilayer film. While this method is attractive for producing well-defined multilayers, corruption of the films may occur via the repetitive immersion and rinsing process and the issue of how this method can be applied to thin film fabrication on any desired substrates is yet to be addressed. In order to overcome this problem, Leibowitz et al. [313]. suggested an alternative and simpler pathway for the production of thin films on substrates of any type. The procedure involved a one-step, exchange–crosslinking–precipitation route, where ligands (that is, alkanethiols) on gold nanoparticles were place-exchanged with , -functionalized dithiols, which in turn caused crosslinking and precipitation of gold nanoparticles on the surface of the substrates. This route may be written as Aun − SR + HSR′ SH → SR′ S − Aun · · · → −&SR′ S − Aun 'm − precipitation in which (HSR′ SH) is the dithiol and (SR) is the alkane thiol. The structural and electronic properties of the thin films produced by this method were found to be comparable with thin films derived from a stepwise layer-by-layer

method. Moreover, since one can produce thin films in a single step using this method, it offers a much simpler route for preparing nanoparticle thin films on almost any substrate.

5.4. Biosensors Current research in biosensor technology is leaning toward the use of gold nanoparticles as an important component of biosensing platforms due their unique optical and electronic properties. Recently, Jia et al. [320]. constructed a horseradish peroxidase (HRP) biosensor using sol–gel processing and gold nanoparticles in a three-dimensional network. In this study, gold nanoparticles were chemisorbed onto thiol groups on the sol–gel network, and HRP was adsorbed onto the surface of gold nanoparticles. Here, gold nanoparticles acted as tiny conduction centers that facilitated the transfer of electrons (produced in the enzymatic reaction where hydrogen peroxide was oxidized) to the electrode surface. Since gold nanoparticles were assembled into a multilayer, the enzyme loading on the biosensor was increased resulting in better signals. Also, both gold nanoparticles and sol–gel had little effect on enzyme activity, imparting the biosensor with high sensitivity and good stability. In 1996, Chen et al. [321] reported that the addition of gold nanoparticles with sizes ranging from 10 to 30 nm to a glucose biosensor resulted in the enhancement of glucose oxidase (GOD) activity. They found that 10 nm gold nanoparticles caused more increase in the GOD activity than 30 nm gold nanoparticles. In their following report, they used even smaller gold nanoparticles (less than 10 nm) to investigate further enhancement in the enzymatic activity. The enhanced activity of GOD was attributed to the quantum size effect, which is related to electron transfer between the nanoparticles and protein molecules, which could elucidate the enhancement of enzyme activity. According to this concept, the surface of metallic clusters is always electron deficient, and the redox potential of nanoparticles is closely related to particle size and is relatively high. When the atoms contained in a particle decrease as the particle size decreases, the redox potential of the nanoparticle increases causing stronger affinity to electrons. Although the enhancement effect of gold nanoparticles on enzyme activity is observed qualitatively and further investigation is necessary, it was an important step for the development of better glucose biosensors. Based on the observations made on the glucose sensor in which gold nanoparticles enhanced the enzymatic activity, it was considered that gold nanoparticles could also be useful for DNA sensors [322–328] considering that gold nanoparticles allow immobilization of biomolecules and could be assembled on gold substrates. The quartz crystal microbalance (QCM) is widely used for DNA sensing and allows real-time measurement of DNA binding and hybridization at the nanogram level. Its potential for DNA hybridization detection has been demonstrated recently. The detection limit of this device could be improved by using DNAfunctionalized gold nanoparticles in a probe/target/probenanoparticle sandwich fashion where the amount of DNA hybridized is increased significantly by introducing DNAfunctionalized gold nanoparticles on the QCM surface. This

44 method causes an amplified frequency shift and substantially extends the sensitivity limits of the QCM detection system. Liu et al. [327] studied the effect of the size of gold nanoparticles on the detection sensitivity enhancement of a QCM system, and they found that the amount of mass change on the QCM sensor increased as the gold nanoparticle size increased until 40 nm and decreased thereafter. Signal enhancement using gold nanoparticles is not just limited to QCM; Lyon et al. [329] utilized gold nanoparticles in the signal enhancement of SPR immunosensing. When gold nanoparticles are incorporated into the SPR biosensing surface such that surface immobilized antibody interacts with antigen–gold nanoparticle conjugate, an increased SPR sensitivity to protein–protein interactions was observed. Picomolar (pM) detection of human IgG has been realized using gold nanoparticle enhancement, with theoretical detection limits for this technique being much lower [328].

6. CONCLUSIONS The rapid growth in the number of applications that colloidal gold has found during the last decade has expanded to new levels. It is likely that this material will be considered more than a simple stain for glass, a marker for TEM studies in molecular biology, or a good model system for colloidal science. Novel applications will be found in the field of nanotechnology and particularly in combination with biological molecules. Ordered assemblies and arrays of these particles may provide materials with novel properties. Supramolecular crystals have been explored and may find novel applications. More importantly, novel technologies will emerge that use this material and exploit its unique optical and electronic properties. One of the fields that will likely encounter fast growth in the near future is the implementation of metallic nanoparticles in fluorescence detection schemes. Nanoparticle alloys of gold and other noble metals such as silver will allow further flexibility in finetuning the optical and electronic properties of this material. Additionally, the optical properties of anisotropic nanoparticles could be implemented to great advantage in surface enhancement of fluorescent emission and Raman scattering. For example, nanorods of this material would concentrate the electric field at the end of the nanorod and this could be exploited to greatly increase the natural fluorescence of biological molecules such as DNA. It is likely that electronic properties such as coupling of SPR modes could be directed at the nanoscale level using assemblies of nanorods for spatially directed transduction of electronic and/or optical excitations. The boom in the number of applications of this material is only likely to continue expanding into more sophisticated applications during the coming years.

Colloidal Gold

Fluorescence The emission of light by a substance after absorbing energy from light of usually shorter wavelength. When an incident light source excites the electrons of a substance these will jump to higher excited levels. When they return to the ground level then energy is released as photons. Fluorescence resonance energy transfer (FRET) The radiationless transfer of excitation energy from a donor to an acceptor. When this transfer occurs the emission of fluorescence by the donor is decreased. The acceptor may or may not be fluorescent. If the acceptor is fluorescent, emission will occur from the acceptor instead of the donor. FRET is a distance-dependent interaction that occurs over distances of 1–10 nm. Monolayer protected clusters (MPCs) Metallic clusters covered by an organic monolayer that prevents their irreversible aggregation. These nanoparticles are most commonly formed by reduction of a noble metal salt in the presence of a surface-active substance that forms a monolayer on the surface of the growing clusters, thus limiting their further growth and stabilizing them against aggregation. Nanoshells Metallic nanoparticles consisting of a metallic shell surrounding a dielectric core. Recently nanoshells have also been prepared that consist of a metallic core surrounded by a metallic shell consisting of a different metal. Self-assembled monolayers Molecular assemblies that are formed spontaneously by immersion of a substrate in to a solution of active surfactant in a solvent. Surface enhanced emission of fluorescence The increased emission of fluorescence resulting when fluorophores interact with the enhanced electromagnetic field around metallic nanoparticles. Surface enhanced Raman spectroscopy (SERS) The dramatic increase in the intensity of an otherwise weak Raman signal produced by rough metallic surfaces. This increase can be as high as 104 –106 or even 108 –1014 for some systems. The main contribution to this enhancement is due to the concentration of the electromagnetic field in the vicinity of rough metallic surfaces. Surface plasmons The free electrons on the surface of a metal that behave like a quasi-free electron gas. They represent the quanta of the oscillations of surface charges and can behave like real surface waves. They can be excited by the electric field of electromagnetic radiation. Surface plasmon resonance (SPR) The condition at which the frequency of the incident electromagnetic radiation couples with that of the oscillating electrons on the surface of a metal.

GLOSSARY Anisometric gold nanoparticles Nonspherical gold nanoparticles. Deoxyribonucleic acid (DNA) A macromolecule consisting of one or two strands of linked deoxyribonucleotides. DNA hybridization Formation of the double helical structure of DNA by pairing of two complementary single strands of DNA.

ACKNOWLEDGMENTS Financial assistance from the Department of Chemical and Environmental Engineering, Armour College of Engineering and the Educational Initiative Research Fund at the Graduate College of the Illinois Institute of Technology is gratefully acknowledged.

Colloidal Gold

REFERENCES 1. L. Lee, G. Seddon, and F. Stephens, “Stained Glass.” Crown, New York, 1976. 2. G. Savage, “Glass and Glassware.” Octopus Books, London, 1975. 3. H. B. Wesier, “Inorganic Colloidal Chemistry,” Vol. 1. Wiley, New York, 1933. 4. F. Mehlman, “Phaidon Guide to Glass.” Prentice Hall, Englewood Cliffs, NJ, 1983. 5. N. H. Moore, “Old Glass.” Tudor, New York, 1935. 6. M. Faraday, Philos. Trans. Roy. Soc. London 147, 145 (1857). 7. L. P. Williams, Ed., “The Selected Correspondence of Michael Faraday,” Vol. 2. Cambridge Univ. Press, London, 1971. 8. W. Ostwald, “Practical Colloid Chemistry,” 4th ed. Dutton, New York, 1924. 9. G. Frens, Nature (London) 241, 20 (1973). 10. M. A. Hayat, Ed., “Colloidal Gold: Principles, Methods, and Applications,” Vols. 1, 2, and 3. Academic Press, San Diego, 1989. 11. L. C. Javois, Ed., “Immunocytochemical Methods and Protocols.” Humana Press, Clifton, NJ, 1999. 12. J. E. Beesley, “Colloidal Gold: New Perspective for Cytochemical Marking.” Oxford Univ. Press, Oxford, UK, 1999. 13. M. Bendayan, “Colloidal Gold Post-Embedding Immunocytochemistry,” Vol. 29. Wiley, Stuttgart, 1995. 14. C. A. Foss and D. L. Feldheim, “Metal Nanoparticles: Synthesis, Characterization, and Applications.” Dekker, New York, 2001. 15. J. Schmitt, P. M. Mächtle, D. Eck, H. Möwald, and C. A. Helm, Langmuir 15, 3256 (1999). 16. D. Eck and C. A. Helm, Langmuir 17, 957 (2001). 17. U. Kreibig and M. Vollmer, “Optical Properties of Metal Clusters.” Springer-Verlag, Berlin, 1995. 18. C. F. Bohren and D. R. Huffman, “Absorption and Scattering of Light by Small Particles.” Wiley, New York, 1983. 19. G. Mie, Ann. Phys. 25, 307 (1908). 20. C. Kittel, “Introduction to Solid State Physics,” 2nd ed. Wiley, New York, 1956. 21. P. Mulvaney, Langmuir 12, 788 (1996). 22. “Handbook of Chemistry and Physics.” CRC Press, Cleveland, 1986. 23. M. Quinten and U. Kreibig, Surface Sci. 172, 557 (1986). 24. J. Wiesner and A. Wokaun, Chem. Phys. Lett. 157, 569 (1989). 25. J. J. Storhoff, A. A. Lazarides, R. C. Mucic, C. A. Mirkin, R. L. Letsinger, and G. C. Schatz, J. Am. Chem. Soc. 122, 4640 (2000). 26. K. S. Suslick, Science 247, 1439 (1990). 27. A. Henglein, Langmuir 15, 6738 (1999). 28. A. Henglein and D. Meisel, Langmuir 14, 7392 (1998). 29. J. Belloni, M. Mostafavi, H. Remita, J. L. Marignier, and M. O. Delcourt, New J. Chem. 22, 1239 (1998). 30. M. Treguer, C. de Cointet, H. Remita, J. Khatouri, M. Mostafavi, J. Amblard, J. Belloni, R. de Keyzer, and J. Belloni, J. Phys. Chem. B 102, 4310 (1998). 31. A. Henglein and R. J. Tausch-Treml, J. Colloid Interface Sci. 80, 84 (1981). 32. J. L. Marignier, J. Beloni, M. O. Delcourt, and J. P. Chevalier, Nature 317, 344 (1985). 33. Y. Yonezawa, T. Sato, M. Ohno, and H. J. Hada, Chem. Soc., Faraday Trans. 83, 1559 (1987). 34. K. Torigoe and K. Esumi, Langmiur 8, 59 (1992). 35. T. Sato, S. Kuroda, A. Takami, Y. Yonazawa, and H. Hada, Appl. Organomet. Chem. 5, 261 (1991). 36. K. Kurihara, J. Kizling, P. Stenius, and J. J. Fendler, J. Am. Chem. Soc. 105, 2574 (1983). 37. A. Henglein, Chem. Mater. 10, 444 (1998). 38. S. L. Cumberland and G. F. Strouse, Langmuir 18, 269 (2002). 39. J. W. Slot and H. J. Geuze, Europ. J. Cell Biol. 38, 87 (1985). 40. S. Chen and K. Kimura, Langmuir 15, 1075 (1999). 41. G. T. Hermanson, “Bioconjugate Techniques.” Academic Press, San Diego, 1996.

45 42. J.-A. He, R. Valluzzi, K. Yang, T. Dolukhanyan, C. Sung, J. Kumar, S. K. Tripathy, L. Samuelson, L. Balogh, and D. A. Tomalia, Chem. Mater. 11, 3268 (1999). 43. H. Ishizuka, T. Tano, K. Torigoe, K. Esumi, and K. Meguro, Colloids Surf. 63, 337 (1992). 44. K. Esumi, N. Sato, K. Torigoe, and K. Meguro, J. Colloid Interface Sci. 149, 295 (1992). 45. A. Henglein, J. Phys. Chem. B 104, 2201 (2000). 46. A. Henglein, J. Phys. Chem. B 104, 6683 (2000). 47. D. J. Elliot, D. N. Furlong, F. Grieser, P. Mulvaney, and M. Giersig, Coll. Surf. A 130, 141 (1997). 48. K. R. Brown, D. G. Walter, and M. J. Natan, Chem. Mater. 12, 306 (2000). 49. D. G. Duff, A. Baiker, and P. P. Edwards, Langmuir 9, 2310 (1993). 50. K. R. Brown, L. A. Lyon, A. P. Fox, B. D. Reiss, and M. J. Natan, Chem. Mater. 12, 314 (2000). 51. E. C. Stathis and A. Fabrikanos, Chem. Ind. 47, 860 (1958). 52. V. Privman, D. V. Goia, J. Park, and E. Matijevic, J. Coll. Int. Sci. 213, 36 (1999). 53. H. Hirai, Y. Nakao, and N. Toshima, J. Macromol. Sci.-Chem. A 12 1117 (1978). 54. H. Hirai, J. Macromol. Sci.-Chem. A 13, 633 (1979). 55. H. Hirai, Y. Nakao, and N. Toshima, J. Macromol. Sci.-Chem. A 13, 727 (1979). 56. M. Y. Han, C. H. Quek, W. Huang, C. H. Chew, and L. M. Gan, Chem. Mater. 11, 1144 (1999). 57. L. Longenberger and G. Mills, J. Phys. Chem. 99, 475 (1995). 58. J. Turkevich, P. C. Stevenson, and J. Hillier, Discuss. Faraday Soc. 11, 55 (1951). 59. M. Brust, M. Walker, D. Bethell, D. J. Schiffrin, and R. Whyman, J. Chem. Soc. Chem. Commun. 801 (1994). 60. M. Brust, J. Fink, D. Bethell, D. J. Schiffrin, and C. Kiely, J. Chem. Soc. Chem. Commun. 1655 (1995). 61. A. C. Templeton, W. P. Wuelfing, and R. W. Murray, Acc. Chem. Res. 33, 27 (2000). 62. N. C. Greenham, X. G. Peng, and A. P. Alivisatos, Phys. Rev. B 54, 17628 (1996). 63. V. L. Colvin, M. C. Schlamp, and A. P. Alivisatos, Nature 370, 354 (1994). 64. T. Cassagneau, T. E. Mallouk, and J. H. Fendler, J. Am. Chem. Soc. 120, 7848 (1998). 65. M. Gao, B. Richer, and S. Kirstein, Adv. Mater. 9, 802 (1997). 66. M. Gao, B. Richer, S. Kirstein, and H. Möhwald, J. Phys. Chem. B 102, 4096 (1998). 67. H. Mattoussi, L. H. Radzilowski, B. O. Dabbousi, E. L. Thomas, M. G. Bawendi, and M. F. Rubner, J. Appl. Phys. 83, 7965 (1998). 68. R. Gangopadhyay and A. De, Chem. Mater. 12, 608 (2000). 69. R. Singh, R. P. Tandon, and S. Chandra, J. Appl. Phys. 70, 243 (1991). 70. Y. P. Ting, K. G. Neoh, E. T. Kang, and K. L. Tan, J. Chem. Technol. Biotechnol. 59, 31 (1994). 71. E. K. Miller, C. J. Brabee, H. Neugebauer, A. J. Heeger, and N. S. Sariciftci, Chem. Phys. Lett. 335, 23 (2001). 72. A. Kushida, M. Yamato, C. Konno, A. Kikuchi, Y. Sakurai, and T. Okano, J. Biomed. Mater. Res. 51, 216 (2000). 73. K. Uchida, O. H. Kwon, E. Ito, T. Aoyagi, A. Kikuchi, M. Yamato, and T. Okano, Macromol. Rapid Comm. 21, 169 (2000). 74. L. K. Ista, V. H. Perez-Luna, and G. P. Lopez, Appl. Environ. Microbiol. 65, 1603 (1999). 75. A. B. R. Mayer, Mater. Sci. Eng. C 6, 155 (1998). 76. S. T. Selvan, J. P. Spatz, H.-A. Klok, and M. Moller, Adv. Mater. 10, 132 (1998). 77. J. J. Watkins and T. J. McCarthy, Chem. Mater. 7, 1991 (1995). 78. K. Sayo, S. Deki, and S. Hayashi, Eur. Phys. J. D 9, 429 (1999). 79. J. Lee, V. C. Sundar, J. R. Heine, M. G. Bawendi, and K. F. Jensen, Adv. Mater. 12, 1102 (2000).

46 80. M. K. Corbierre, N. S. Cameron, M. Sutton, S. G. J. Mochrie, L. B. Lurio, A. Ruhm, and R. B. Lennox, J. Am. Chem. Soc. 123 10411 (2001). 81. J. M. G. Cowie, “Polymers: Chemistry and Physics of Modern Materials 2.” Blackie, New York, 1991. 82. J. M. Stouffer and T. J. McCarthy, Macromolecules 21, 1204 (1988). 83. N. C. Greenham, X. G. Peng, and A. P. Alivisatos, Phys. Rev. B. 54, 17628 (1996). 84. Y. Zhou, H. Itoh, T. Uemura, K. Naka, and Y. Chujo, Langmuir 18, 277 (2002). 85. R. Gangopadhyay and A. De, Chem. Mater. 12, 608 (2000). 86. D. I. Svergun, M. B. Kozin, P. V. Konarev, E. V. Shtykova, V. V. Volkov, D. M. Chernyshov, P. M. Valetsky, and L. M. Bronstein, Chem. Mater. 12, 3552 (2000). 87. A. B. Lowe, B. S. Sumerlin, M. S. Donovan, and C. L. McCormick, J. Am. Chem. Soc. 124, 11562 (2002). 88. B. S. Sumerlin, M. S. Donovan, Y. Mitsukami, A. B. Lowe, and C. L. McCormick, Macromolecules 34, 6561 (2001). 89. Y. Mitsukami, M. S. Donovan, A. B. Lowe, and C. L. McCormick, Macromolecules 34, 2248 (2001). 90. M. S. Donovan, A. B. Lowe, B. S. Sumerlin, and C. L. McCormick, Macromolecules 35, 4123 (2002). 91. M. S. Donovan, T. Sanford, A. B. Lowe, Y. Mitsukami, B. S. Sumerlin, and C. L. McCormick, Macromolecules 35, 4570 (2002). 92. M. Tagaki, J. Phys. Soc. Jpn. 9, 359 (1954). 93. D. A. Buffat and J. P. Bonel, Phys. Rev. A 13, 2289 (1976). 94. J. P. Wilcoxon, J. E. Martin, F. Parsapour, B. Wiedenman, and D. F. Kelley, J. Chem. Phys. 108, 9137 (1998). 95. M. Haruta, S. Tsubota, T. Kobayashi, H. Kayegama, M. J. Genet, and B. Delmon, J. Catal. 144, 175 (1993). 96. H. Sakurat and M. Haruta, Appl. Catal. A: General 127, 93 (1995). 97. C. Petit, P. Lixon, and M. P. Pileni, J. Phys. Chem. 97, 12974 (1993). 98. R. A. Caruso, M. Ashokkumar, and F. Grieser, Langmuir 18, 7831 (2002). 99. A. Henglein, J. Phys. Chem. 97, 5457 (1993). 100. F. Mafune, J.-Y. Kohno, Y. Takeda, and T. Kondow, J. Phys. Chem. B 105, 9050 (2001). 101. F. Mafune, J.-Y. Kohno, Y. Takeda, and T. Kondow, J. Phys. Chem. B 106, 8555 (2002). 102. H. Kurita, A. Takami, and S. Koda, Appl. Phys. Lett. 72, 789 (1998). 103. F. Mafune, J.-Y. Kohno, Y. Takeda, T. Kondow, and H. Sawabe, J. Phys. Chem. B 105, 5114 (2001). 104. H. Eckstein and U. Kreibig, Z. Phys. D 26, 239 (1993). 105. Y. Takeuchi, T. Ida, and K. Kimura, J. Phys. Chem. B 101, 1322 (1997). 106. A. Takami, H. Kurita, and S. Koda, J. Phys. Chem. B 103, 1226 (1999). 107. K. Okitsu, A. Yue, S. Tanabe, H. Matsumoto, and Y. Yobiko, Langmuir 17, 7717 (2001). 108. W. Chen, W. Cai, L. Zhang, G. Wang, and L. Zhang, J. Coll. Int. Sci. 238, 291 (2001). 109. Y. Nagata, Y. Mizukoshi, and K. Okitsu, Radiation Res. 146, 333 (1996). 110. Y. Zhou, C. Y. Wang, Y. R. Zhu, and Z. Y. Chen, Chem. Mater. 11, 2310 (1999). 111. W. O. Milligan and R. H. Morriss, J. Am. Chem. Soc. 86, 3461 (1964). 112. K. Esumi, K. Matsuhisa, and K. Torigoe, Langmuir 11, 3285 (1995). 113. S. Link, M. B. Mohamed, and M. A. El-Sayed, J. Phys. Chem. B 103, 3073 (1999). 114. C. A. Foss, G. L. Hornyak, J. A. Stockert, and C. R. Martin, J. Phys. Chem. 96, 7497 (1992). 115. C. A. Foss Jr., G. L. Hornyak, J. A. Stockert, and C. R. Martin, J. Phys. Chem. B 98, 2963 (1994). 116. G. L. Hornyak, C. J. Patrissi, and C. R. Martin, J. Phys. Chem. B 101, 1548 (1997).

Colloidal Gold 117. G. L. Hornyak and C. R. Martin, Thin Solid Films 303, 84 (1997). 118. M. L. Sandrock, C. D. Pibel, F. M. Geiger, and C. A. Foss Jr., J. Phys. Chem. B 103, 2668 (1999). 119. C. R. Martin, Science 266, 1961 (1994). 120. G. L. Hornyak, C. J. Patrissi, and C. R. Martin, J. Phys. Chem. B 101, 1548 (1997). 121. J. C. Hulteen, C. J. Patrissi, D. L. Miner, E. R. Crosthwait, E. B. Oberhauser, and C. R. Martin, J. Phys. Chem. B 101, 7727 (1997). 122. C. A. Foss, G. L. Hornyak, M. J. Tierney, and C. R. Martin, J. Phys. Chem. 96, 9001 (1992). 123. C. K. Preston and M. Moskovits, J. Phys. Chem. 92, 2957 (1988). 124. C. K. Preston and M. Moskovits, J. Phys. Chem. 97, 8405 (1993). 125. M. J. Tierney and C. R. Martin, J. Phys. Chem. 93, 2878 (1989). 126. Y.-Y. Yu, S.-S. Chang, C.-L Lee, and C. R. C. Wang, J. Phys. Chem. B 101, 6661 (1997). 127. A. H. Lu, G. H. Lu, A. M. Kessinger, and C. A. Foss Jr., J. Phys. Chem. B 101, 9139 (1997). 128. P. F. Liao, R. K. Chang, and T. E. Furtak, “Surface Enhanced Raman Scattering.” Plenum Press, New York, 1982. 129. M. C. Buncick, R. J. Warmack, and T. C. Ferrell, J. Opt. Soc. Am. B 4, 927 (1988). 130. C. Schönenberger, B. M. I. Van der Zande, L. G. J. Fokkink, M. Henny, C. Schmid, M. Krüger, A. Bachtold, R. Huber, H. Birk, and U. Staufer, J. Phys. Chem. B 101, 5497 (1997). 131. N. R. Jana, L. Gearheart, and C. J. Murphy, J. Phys. Chem. B 105, 4065 (2001). 132. N. R. Jana, L. Gearheart, S. O. Obare, and C. J. Murphy, Langmuir 18, 922 (2002). 133. J. C. Wittman and P. Smith, Nature 352, 414 (1991). 134. R. E. Gill, G. Hadziioannou, P. Lang, F. Garnier, and J. C. Wittman, Adv. Mater. 9, 331 (1997). 135. P. Hubert, H. Dreyfus, D. Guillon, and Y. Galerne, J. Phys. II 5, 1371 (1995). 136. S. Ravaine, G. E. Fanucci, C. T. Seip, J. H. Adair, and D. R. Talham, Langmuir 14, 708 (1998). 137. L. D. Marks and D. J. Smith, J. Crystal Growth 54, 425 (1981). 138. D. J. Smith and L. D. Marks, J. Crystal Growth 54, 433 (1981). 139. R. Gans, Ann. Phys. 47, 270 (1915). 140. D. C. Skillman and C. R. Berry, J. Chem. Phys. 48, 3297 (1968). 141. M. B. Mohamed, Z. L. Wang, and M. A. El-Sayed, J. Phys. Chem. A 103, 10255 (1999). 142. M. A. El-Sayed, Acc. Chem. Res. 34, 257 (2001). 143. R. D. Averitt, D. Sarkar, and N. Halas, J. Phys. Rev. Lett. 78, 4217 (1997). 144. S. J. Oldenburg, R. D. Averitt, S. L. Westcott, and N. J. Halas, J. Chem. Phys. Lett. 288, 243 (1998). 145. R. D. Averitt, S.L. Westcott, and N. J. Halas, J. Phys. Rev. B 58, 10203 (1998). 146. S. J. Oldenburg, S. L. Westcott, R. D. Averitt, and N. Halas, J. Chem. Phys. 111, 4729 (1999). 147. R. D. Averitt, S. L. Westcott, and N. J. Halas, J. Opt. Soc. Am. B 16, 1824 (1999). 148. S. J. Oldenburg, J. B. Jackson, S. L. Westcott, and N. J. Halas, Appl. Phys. Lett. 75, 2897 (1999). 149. S. R. Sershen, S. L. Westcott, N. J. Halas, and J. L. West, J. Biomed. Mater. Res. 51, 293 (2000). 150. G. D. Hale, J. B. Jackson, O. E. Shmakova, T. R. Lee, and N. J. Halas, Appl. Phys. Lett. 78, 1502 (2001). 151. J. B. Jackson and N. J. Halas, J. Phys. Chem. B 105, 2743 (2001). 152. D. D. Smith, L.A. Snow, L. Sibille, and E. Ignont, J. Non-Cryst. Solids 285, 256 (2001). 153. V. C. Coffey, Laser Focus World 37, 46 (2001). 154. J. J. Diao and G. D. Chen, J. Phys. D 34, L79, (2001). 155. M. C. Netti, S. Coyle, J. J. Baumberg, M. A. Ghanem, P. R. Birkin, P. N. Bartlett, and D. M. Whittaker, Adv. Mater. 13, 1368 (2001). 156. S. Lal, S. L. Westcott, R. N. Taylor, J. B. Jackson, P. Nordlander, and N. J. Halas, J. Phys. Chem. B 106, 5609 (2002).

Colloidal Gold 157. T. Pham, J. B. Jackson, N. J. Halas, and T. R. Lee, Langmuir 18, 4915 (2002). 158. E. Prodan, A. Lee, and P. Nordlander, Chem. Phys. Lett. 360, 325 (2002). 159. Y. G. Sun and Y. N. Xia, Anal. Chem. 74, 5297 (2002). 160. T. J. Beveridge and R. G. E. Murray, J. Bacteriol. 141, 876, (1980). 161. T. J. Beveridge and R. J. Doyle, “Metal Ions and Bacteria.” Wiley, New York, 1989. 162. G. Southam and T. J. Beveridge, Geochim. Cosmochim. Acta 60, 4369 (1996). 163. D. Fortin and T. J. Beveridge in “Biomineralization; From Biology to Biotechnology and Medical Applications” (E. Baeuerlein, Ed.), p. 7. Wiley–VCH, Weinheim, 2000. 164. P. Mukherjee, A. Ahmad, D. Mandal, S. Senapati, S. R. Sainkar, M. I. Khan, R. Ramani, R. Parischa, P. V. Ajayakumar, M. Alam, M. Sastry, and R. Kumar, Angew. Chem. Int. Ed. 40, 3585 (2001). 165. P. Mukherjee, S. Senapati, D. Mandal. A. Ahmad, M. I. Khan, R. Kumar, and M. Sastry, Chem. Biochem. 5, 461 (2002). 166. N. Braun, W. Meining, U. Hars, M. Fischer, R. Ladenstein, R. Huber, A. Bacher, S. Weinkauf, and L. Bachmann, J. Mol. Biol. 321, 341 (2002). 167. Z. Zhang and M. G. Lagally, Science 76, 377 (1997). 168. C. S. Weisbecker, M. V. Merritt, and G. M. Whitesides, Langmuir 12, 3763 (1996). 169. K. S. Mayya, V. Patil, and M. Sastry, Langmuir 13, 3944 (1997). 170. J. E. Houston and H. I. Kim, Acc. Chem. Res. 35, 547 (2002). 171. N. K. Chaki and K. Vijayamohanan, Biosens. Bioelectr. 17, 1 (2002). 172. C. M. Knobler and D. K. Schwartz, Curr. Opinion. Coll. Interf. Sci. 4, 46 (1999). 173. M. J. Hostetler and R. W. Murray, Curr. Opinion. Coll. Interf. Sci. 2, 42 (1997). 174. A. R. Bishop and R. G. Nuzzo, Curr. Opinion. Coll. Interf. Sci. 1, 127 (1996). 175. A. Ulman, Chem. Rev. 96, 1533 (1996). 176. M. Mrksich and G. M. Whitesides, Ann. Rev. Biophys. Biomol. Struct. 25, 55 (1996). 177. A. Ulman, “An Introduction to Ultrathin Organic Films: From Langmuir–Blodgett to Self Assembly.” Academic Press, Boston, 1991. 178. S. J. Green, J. J. Stokes, M. J. Hostetler, J. Pietron, and R. W. Murray, J. Phys. Chem. B 101, 2663 (1997). 179. M. J. Hostetler, J. E. Wingate, C.-J. Zhong, J. E. Harris, R. W. Vachet, M. R. Clark, J. D. Londono, S. J. Green, J. J. Stokes, G. D. Wignall, G. L. Glish, M. D. Porter, N. D. Evans, and R. W. Murray, Langmuir 14, 17 (1998). 180. R. H. Terrill, T. A. Postlethwaite, C. Chen, C. Poon, A. Terzis, A. Chen, J. E. Hutchison, M. R. Clark, G. W., J. D. Londono, D. Superfine, M. Falvo, C. S. Johnson, E. T. Samulski, and R .W. Murray, J. Am. Chem. Soc. 117, 12537 (1995). 181. T. G. Schaaff, G. Knight, M. N. Shafigullin, R. F. Borkman, and R. L. Whetten, J. Phys. Chem. B 102, 10643 (1998). 182. A. C. Templeton, S. Chen, S. M. Gross, and R. W. Murray, Langmuir 15, 66 (1999). 183. W. P. Wuelfing, S. M. Gross, D. T. Miles, and R. W. Murray, J. Am. Chem. Soc. 120, 12696 (1998). 184. M. Brust, J. Fink, D. Bethell, D. J. Schiffrin, and C. Kiely, J. Chem. Soc., Chem. Commun. 1655 (1995). 185. S. R. Johnson, S. D. Evans, S. W. Mahon, and A. Ulman, Langmuir 13, 51 (1997). 186. S. Chen and R. W. Murray, Langmuir 15, 682 (1999). 187. P. A. Buining, B. M. Humbel, A. P. Philipse, and A. J. Verkleij, Langmuir 13, 3921 (1997). 188. M. J. Hostetler, S. J. Green, J. J. Stokes, and R. W. Murray, J. Am. Chem. Soc. 118, 4212 (1996). 189. S. R. Johnson, S. D. Evans, and R. Brydson, Langmuir 14, 6639 (1998).

47 190. L. A. Porter, Jr., D. Ji, S. L. Westcott, M. Graupe, R. S. Czernuszewicz, N. J. Halas, and T. R. Lee, Langmuir 14, 7378 (1998). 191. Y. S. Shon, C. Mazzitelli, and R. W. Murray, Langmuir 17, 7735 (2001). 192. D. E. Cliffel, F. P. Zamborini, S. M. Gross, and R. W. Murray, Langmuir 16, 9699 (2000). 193. A. Aguila and R. W. Murray, Langmuir 16, 5949 (2000). 194. A. C. Templeton, S. Chen, S. M. Gross, and R. W. Murray, J. Phys. Chem. B 104, 564 (2000). 195. N. Kometani, M. Tsubonishi, T. Fujita, K. Asami, and Y. Yonezawa, Langmuir 17, 578 (2001). 196. C. Yee, M. Scotti, A. Ulman, H. White, M. Rafailovich, and J. Sokolov, Langmuir 15, 4314 (1999). 197. H. Imahori, M. Arimura, T. Hanada, Y. Nishimura, I. Yamazaki, Y. Sakata, and S. Fukuzumi, J. Am. Chem. Soc. 123, 335 (2001). 198. M. J. Hostetler, A. C. Templeton, and R. W. Murray, Langmuir 15, 3782 (1999). 199. R. S. Ingram, M. J. Hostetler, and R. W. Murray, J. Am. Chem. Soc. 119, 9175 (1997). 200. M. G. Warner, S. M. Reed, and J. E. Hutchison, Chem. Mater. 12, 3316 (2000). 201. P. Ionita, A. Caragheorgheopol, B. C. Gilbert, and V. Chechik, J. Am. Chem. Soc. 124, 9048 (2002). 202. S. J. Green, J. J. Pietron, J. J. Stokes, M. J. Hostetler, H. Vu, W. P. Wuelfing, and R. W. Murray, Langmuir 144, 5612 (1998). 203. M. S. Fleming and D. R. Walt, Langmuir 17, 4836 (2001). 204. J. B. Schlenoff, M. Li, and H. Ly, J. Am. Chem. Soc. 117, 12528 (1995). 205. M. Sastry, N. Lala, V. Patil, S. P. Chavan, and A. G. Chittiboyina, Langmuir 14, 4138 (1998). 206. J. J. Storhoff, R. Elghanian, R. C. Mucic, C. A. Mirkin, and R. L. Letsinger, J. Am. Chem. Soc. 120, 1959 (1998). 207. S. Connolly, S. Cobbe, and D. Fitzmaurice, J. Phys. Chem. B 105, 2222 (2001). 208. K. Aslan and V. H. Pérez-Luna, Langmuir 18, 6059 (2002). 209. N. Lala, A. G. Chittiboyina, and S. P. Chavan, Coll. Surf. A 205, 15 (2002). 210. M. Sastry, K. S. Mayya, and V. Patil, Langmuir 14, 5921 (1998). 211. V. Patil and M. Sastry, Langmuir 14, 2707 (1998). 212. V. Patil, K. S. Mayya, S. D. Pradhan, and M. Sastry, J. Am. Chem. Soc. 119, 9281 (1997). 213. E. De Harven, R. Leung, and H. Christensen, J. Cell Biol. 99, 53 (1984). 214. W. Faulk and G. Taylor, Immunochemistry 8, 1081 (1979). 215. E. Romano, C. Stolinsky, and N. Hugh-Jones, Immunochemistry 11, 521 (1974). 216. J. Roth, M. Bendayan, and L Orci, J. Histochem. Cytochem. 28, 55 (1980). 217. S. Eskelinen and R. Peura, Scan. Electr. Microsc. 2, 1765 (1988). 218. R. M. Albrecht and G. M. Hodges, Eds., “Biotechnology and Bioapplications of Colloidal Gold.” Scanning Microscopy International, IL, 1988. 219. M. M. Kay, in “Principles and Techniques of Scanning Electron Microscopy” (M. A. Hayat, Ed.), Vol. 6, p. 338. Van Nostrand Reinhold, New York, 1978. 220. R. S. Molday and P. Maher, J. Histochem. 12, 273 (1980). 221. E. Skutelsky and J. Roth, J. Histochem. Cytochem. 34, 693 (1986). 222. A. W. Vorbrodt, J. Histochem. Cytochem. 35, 1261 (1987). 223. N. Thurauf, R. Dermietzel, and P. Kalweit, J. Ultrastruct Res. 84, 103 (1983). 224. J. Beesley, “Royal Microscopical Society Handbook,” Vol. 17. Oxford Univ. Press, London, 1989. 225. T. Wenzel and W. Baumeister, Nat. Struct. Biol. 2, 199 (1995). 226. A. P. Alivisatos, K. P. Johnson, X. Peng, T. E. Wilson, C. J. Loweth, M. P. Bruchez Jr., and P. G. Schultz, Nature 382, 609 (1996).

48 227. J. F. Hainfield, in “Proc. 54th Ann. Mtg., Micr. Soc. Amer.” (G. W. Bailey, J. M. Corbett, R. V. W. Dimlich, J. R. Michael, N. J. Zaluzec, Eds.), p. 898. San Francisco Press, San Francisco, 1996. 228. C. Mangeney, F. Ferrage, I. Aujard, V. Marchi-Artzner, L. Jullien, O. Ouari, E. D. Rekai, A. Laschewsky, I. Vikholm, and J. W. Sadowski, J. Am. Chem. Soc. 124, 5811 (2002). 229. A. Laschewsky, O. Quari, C. Mangeney, and L. Jullien, Macromol. Symp. 164, 323 (2001). 230. R. Jordan, A. Ulman, J. F. Kang, M. H. Rafailovich, and J. Sokolov, J. Am. Chem. Soc. 121, 1016 (1999). 231. R. Jordan, N. West, A. Ulman, Y.-M. Chou, and O. Nuyken, Macromolecules 34, 1606 (2001). 232. R. Jordan and A. Ulman, J. Am. Chem. Soc. 120, 243 (1998). 233. R. Jordan, A. Ulman, J. F. Kang, M. Rafailovich, and J. Sokolov, J. Am. Chem. Soc. 121, 1016 (1999). 234. R. Jordan, K. Martin, H. J. Rader, and K. K. Unger, Macromolecules 34, 8858 (2001). 235. N. Nath and A. Chilkoti, J. Am. Chem. Soc. 123, 8197 (2001). 236. D. A. Tomalia, Angew. Chem. Int. Ed. Engl. 29, 138 (1990). 237. F. Zeng and S. C. Zimmerman, Chem. Rev. 97, 1681 (1997). 238. D. A. Tomalia, A. M. Naylor, and W. A. Goddard, Angew. Chem. 102, 119 (1990). 239. D. A. Tomalia and H. D. Durst, Top. Curr. Chem. 165, 193 (1993). 240. G. R. Newkome, C. N. Moorefield, and F. Vögtle, “Dendritic Macromolecules: Concepts, Syntheses, Perspectives.” VCH, Weinheim, 1996. 241. J. M. J. Fréchet, Science 263, 1710 (1994). 242. B. I. Voit, Acta Polym. 46, 87 (1995). 243. N. Ardoin and D. Astruc, Bull. Soc. Chim. Fr. 132, 875 (1995). 244. M. Zhao, H. Tokuhisa, and R. M. Crooks, Angew. Chem. Int. Ed. Engl. 36, 2596 (1997). 245. M. F. Ottaviani, F. Montalti, N. J. Turro, and D. A. Tomalia, J. Phys. Chem. B 101, 158 (1997). 246. K. Esumi, A. Suzuki, N. Aihara, K. Usui, and K. Torigoe, Langmuir 14, 3157 (1998). 247. M. E. Garcia, L. A. Baker, and R. M. Crooks, Anal. Chem. 71, 256 (1999). 248. L. Balogh, R. Valluzzi, K. S. Laverdure, S. P. Gido, G. L. Hagnauer, and D. A Tomalia, J. Nanopart. Res. 1, 353 (1999). 249. K. Esumi, A. Suzuki, A. Yamahira, and K. Torigoe, Langmuir 16, 2604 (2000). 250. F. Grohn, B. J. Bauer, Y. A. Akpalu, C. L. Jackson, and E. J. Amis, Macromolecules 13, 6042 (2000). 251. J.-A. He, R. Valluzzi, K. Yang, T. Dolukhanyan, C. Sung, J. Kumar, S. K. Tripathy, L. Samuelson, L. Balogh, and D. A. Tomalia, Chem. Mater. 11, 3268 (1999). 252. K. Aoi, K. Itoh, and M. Okada, Macromolecules 28, 5391 (1995). 253. K. Esumi, T. Hosoya, A. Suzuki, and K. Torigoe, Langmuir 16, 2978 (2000). 254. A. Manna, T. Imae, K. Aoi, M Okada, and T. Yogo, Chem. Mater. 13, 1674 (2001). 255. V. Chechik and R. M. Crooks, Langmuir 15, 6364 (1999). 256. L. M. Liz-Marzan, M. Giersig, and P. Mulvaney, Langmuir 12, 4329 (1996). 257. S. T. Selvan, T. Hayakawa, M. Nogami, Y. Kobayashi, L. M. LizMarzan, Y. Hamanaka, and A. Nakamura, J. Phys. Chem. B, 106, 10157 (2002). 258. F. Garcia-Santamaria, V. Salgueirino-Maceira, C. Lopez, and L. M. Liz-Marzan, Langmuir 18, 4519 (2002). 259. Y. Kobayashi, M. A. Correa-Duarte, and L. M. Liz-Marzan, Langmuir 17, 6375 (2001). 260. S. Bharathi and O. Lev, Chem. Comm. 23, 2303 (1997). 261. S. Bharathi, N. Fishelson, and O. Lev, Langmuir 15, 1929 (1999). 262. C. J. Brinker and G. W. Scherer, “Sol–Gel Science.” Academic Press, New York, 1990. 263. A. Martino, S. A. Yamanaka, J. S. Kowala, and D. L. Loy, Chem. Mater. 9, 423 (1997).

Colloidal Gold 264. K. C. Grabar, R. G. Freeman, M. B. Hommer, and M. J. Natan, Anal. Chem. 67, 735 (1995). 265. K. C. Grabar, P. C. Smith, M. D. Musick, J. A. Davis, D. G. Walter, M. A. Jackson, A. P. Guthrie, and M. J. Natan, J. Am. Chem. Soc. 118, 1148 (1996). 266. K. C. Grabar, K. J. Allison, B. E. Baker, R. M. Bright, K. R. Brown, R. G. Freeman, A. P. Fox, C. D. Keating, M. D. Musick, and M. J. Natan, Langmuir 12, 2353 (1996). 267. P. Mukherjee, A. Ahmad, D. Mandal, S. Senapati, S. R. Sainkar, M. I. Khan, R. Ramani, R. Parischa, P. V. Ajayakumar, M. Alam, M. Sastry, and R. Kumar, Angew. Chem. Int. Ed. 40, 3585 (2001). 268. D. V. Leff, L. Brandt, and J. R. Heath, Langmuir 12, 4723 (1996). 269. B. D. Ratner and D. G. Castner, Electron spectroscopy for chemical analysis, in “Surface Analysis–Techniques and Applications” (J. C. Vickerman, Ed.), pp. 43–98. Wiley, Chichester, UK, 1997. 270. J. A. Harnisch, A. D. Pris, and M. D. Porter, J. Am. Chem. Soc. 123, 5829 (2001). 271. O. Kohlmann, W. E. Steinmetz, X-A, Mao, W. P. Wuelfing, A .C. Templeton, R. W. Murray, and C. S. Johnson, J. Phys. Chem. B 105, 8801 (2001). 272. M. J. Hostetler, J. J. Stokes, and R. W. Murray, Langmuir 12, 3604 (1996). 273. E. W. L. Chan and L. Yu, Langmuir 18, 311 (2002). 274. F. X. Zhang, W. Zheng, M. M. Maye, Y. Lou, L. Han, and C.-J. Zhong, Langmuir 16, 9639 (2000). 275. M. A. Bryant and J. E. Pemberton, J. Am. Chem. Soc. 113, 8284 (1991). 276. P. C. Lee and D. Meisel, J. Phys. Chem. 86, 3391 (1982). 277. H. Xu, C.-H. Tseng, T. J. Vickers, C. K. Mann, and J. B. Schlenoff, Surf. Sci. 311, L707 (1994). 278. C. J. Sandroff, S. Garoff, and K. P. Leung, Chem. Phys. Lett. 96, 547 (1983). 279. C. A. Mirkin, R. L. Letsinger, R. C. Mucic, and J. J. Storhoff, Nature 382, 607 (1996). 280. L. J. Kricka, “Nonisotopic DNA Probe Techniques.” Academic Press, San Diego, 1992. 281. B. D. Hames and S. J. Higgins, “Gene Probes 1.” IRL Press, New York, 1995. 282. J. Wang, E. Palecek, P. E. Nielsen, G. Rivas, X. Cai, H. Shiraishi, N. Dontha, D. Luo, and P. A. M Farias, J. Am. Chem. Soc. 118, 7667 (1996). 283. S. Tyagi and F. R Kramer, Nat. Biotechnol. 14, 303 (1996). 284. R. Elghanian, J. J. Storhoff, R. C. Mucic, R. L. Letsinger, and C. A. Mirkin, Science 277, 1078 (1997). 285. R. A. Reynolds, III, C. A. Mirkin, and R. L. Letsinger, J. Am. Chem. Soc. 122, 3795 (2000). 286. T. A. Taton, G. Lu, and C. A. Mirkin, J. Am. Chem. Soc. 123, 5164 (2001). 287. Y. Cao, R. Jin, and C. A. Mirkin, J. Am. Chem. Soc. 123, 7961 (2001). 288. H. Schmitt, A. Badia, L. Dickson, L. Reven, and R. B. Lennox, Adv. Mater. 10, 475 (1998). 289. C. Templeton, F. P. Zamborini, W. P. Wuelfing, and R. W. Murray, Langmuir 16, 6682 (2000). 290. J. Simard, C. Briggs, A. K. Boal, and V. M. Rotello, Chem. Commun. 1943 (2000). 291. F. P. Zamborini, J. F. Hicks, and R. W. Murray, J. Am. Chem. Soc. 122, 4514 (2000). 292. W. P. Wuelfing, F. P. Zamborini, A. C. Templeton, X. Wen, H. Yoon, and R. W. Murray, Chem. Mater. 13, 87 (2001). 293. Y. Kim, R. C. Johnson, and J. T. Hupp, Nano Lett. 1, 165 (2001). 294. S. Berchmans, P. J. Thomas, and C. N. R. Rao, J. Phys. Chem. B 106, 4647 (2002). 295. T. Morris, H. Copeland, E. McLinden, S. Wilson, and G. Szulczewski, Langmuir 18, 7261 (2002). 296. P. Avouris and B. N. J. Persson, J. Phys. Chem. 88, 837 (1984).

Colloidal Gold 297. B. N. J. Persson and S. Anderson, Phys. Rev. B 29, 4382 (1984). 298. G. Cnossen, K. E. Drabe, and D. A. Wiersma, J. Chem. Phys. 98, 5276 (1993). 299. J. R. Lakowicz, Anal. Biochem. 298, 1 (2001). 300. K. Sokolov, G. Chumanov, and T. M. Cotton, Anal. Chem. 70, 3898 (1998). 301. T. Liebermann and W. Knoll, Colloids Surf. A 171, 115 (2000). 302. I. Gryczynski, J. Malicka, Y. Shen, Z. Gryczynski, and J. R. Lakowicz, J. Phys. Chem. B 106, 2191 (2002). 303. D. J. Maxwell, J. R. Taylor, and S. Nie, J. Am. Chem. Soc. 124, 9606 (2002). 304. G. Wang, J. Zhang, and R. W. Murray, Anal. Chem. 74, 4320 (2002). 305. P. Hindebrandt and M. Stockburger, J. Phys. Chem. 88, 5935 (1984). 306. A. C. Pineda and D. Ronis, J. Chem. Phys. 83, 5330 (1985). 307. I. Pockrand, A. Brillante, and D. Mobius, Chem. Phys. Lett. 69, 499 (1980). 308. R. G. Freeman, K. C. Grabar, K. J. Allison, R. M. Bright, J. A. Davis, A. P. Guthrie, M. B. Hommer, M. A. Jackson, P. C. Smith, D. G. Walter, and M. J. Natan, Science 267, 1629 (1995). 309. M. D. Musick, D. J. Pena, S. L. Botsko, T. M. McEvoy, J. N. Richardson, and M. J. Natan, Langmuir 15, 844 (1999). 310. K. C. Grabar, K. R. Brown, C. D. Keating, S. J. S., S.-L. Tang, and M. J. Natan, Anal Chem. 69, 471 (1997). 311. N. Nath and A. Chilkoti, Anal. Chem. 74, 504 (2002). 312. H. X. He, H. Zhang, Q. G. Li, T. Zhu, S. F. Y. Li, and Z. F. Liu, Langmuir 16, 3846 (2000). 313. F. L. Leibowitz, W. Zheng, M. M. Maye, and C.-J. Zhong, Anal. Chem. 71, 5076 (1999). 314. L. Scheibler, P. Dumy, M. Boncheva, K. Leufgen, H. J. Mathieu, M. Mutter, and H. Vogel, Angew. Chem. 38, 696 (1999).

49 315. M. D. Musick, C. D. Keating, L. A. Lyon, S. L. Botsko, D. J. Pena, W. D. Holliway, T. M. McEvoy, J. N. Richardson, and M. J. Natan, Chem. Mater 12, 2869 (2000). 316. J. F. Hicks, Y. Seok-Shon, and R. W. Murray, Langmuir 18, 2288 (2002). 317. J. Schmitt, G. Decher, W. J. Dressick, S. L. Brandow, R. E. Geer, S. Shashidhar, and J. M. Calvert, Adv. Mater. 9, 61 (1997). 318. S. Pathak, M. T. Greci, R. C. Kwong, K. Mercado, G. K. S. Prakash, G. A. Olah, and M. E. Thompson, Chem. Mater. 12, 1985 (2000). 319. M. Brust, D. Bethell, C. J. Kiely, and D. J. Schiffrin, Langmuir, 14, 5425 (1998). 320. J. Jia, B. Wang, A. Wu, G. Cheng, Z. Li, and S. Dong, Anal. Chem. 74, 2217 (2002). 321. Z. J. Chen, X. M. Ou, F. Q. Tang, L. Jiang, Colloids and Surfaces B 7, 173 (1996). 322. L. Lin, H. Zhao, J. Li, J. Tang, M. Duan, and L. Jiang, Biochem. Biophys. Res. Comm. 274, 817 (2000). 323. F. Patolsky, K. T. Ranjit, A. Lichtenstein, and I. Willner, Chem. Comm. 12, 1025 (2000). 324. I. Willner, F. Patolsky, Y. Weizmann, and B. Willner, Talanta 56, 847 (2002). 325. X. C. Zhou, S. J. O’Shea, and S. F. Y. Li, Chem. Comm. 11, 953 (2000). 326. T. Liu, J. Tang, and L. Jiang, Biochem. Biophys. Res. Comm. 295, 14 (2002). 327. T. Liu, J. Tang, H. Zhao, Y. Deng, and L. Jiang, Langmuir 18, 5624 (2002). 328. H.-Y. Gu, A.-M. Yu, S.-S. Yuan, and H.-Y. Chen, Anal. Lett. 35, 647 (2002). 329. L. A. Lyon, M. D. Musick, and M. J. Natan, Anal. Chem. 70, 51774 (1998).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Computational Atomic Nanodesign Michael Rieth, Wolfram Schommers Forschungszentrum Karlsruhe, Karlsruhe, Germany

CONTENTS 1. Introduction 2. Quantum-Theoretical Treatment of Nanosystems 3. Basic Properties at the Nanolevel 4. Modeling of Nanosystems 5. Interaction Potentials 6. Molecular Dynamics 7. Monte Carlo Method 8. Summary Glossary References

1. INTRODUCTION Only a few years ago technology had mainly to do with macroscopic systems and aggregates with micrometer-scale properties. The situation changed fundamentally with the rise of nanotechnology; this technology will have an essential influence on the development of new devices in all or almost all fields of technology, medicine, biology etc., and these developments will influence our future life considerably. It is well known that nanosystems can have properties which are distinctly different from those of macroscopic systems: properties which are well defined and clearly fixed in the bulk are very often no longer typical features in nanophysics and nanotechnology. If the particle number is decreased so that the size of the system is in the nanometer region, new effects emerge which have no counterpart in other fields of physics, chemistry, and technology. Let us give an example. The melting temperature of a macroscopic system is usually well defined, and the system is thermally stable up to the melting point. Of course there can be phase transitions but the various phases define stable configurations. In contrast to macroscopic systems, the melting temperature of a nanosystem in general depends on the particle number and is also a function of the outer shape of the system. Moreover, the melting temperature of certain ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

nanosystems is not defined and not clearly fixed, which can be explained by the faster occurrence of sublimation compared to the melting process. This is a completely new situation, in particular in connection with materials research. Clearly, the construction of nanomachines and other aggregates is based on components whose material properties have an essential influence on the function on such systems. Therefore, the definition and classification of material properties at the nanolevel is a particular challenge and is just at the beginning. Within such a program not only do further sophisticated experiments have to be prepared, but also specific methods for the theoretical treatment of such systems have to be selected and refined carefully. Nanotechnology will also lead to drastic new insights in connection with biological systems, not only by the manipulation of the features of existing biological structures, but also with respect to questions as “What is a biological system?” or, quite generally “What is life?” It is usually assumed that biological features are strongly correlated with systems of sufficiently large complexity, in particular, in connection with complex molecules. But do we really need complex systems and complex organic molecules, respectively, for all phenomena which arise in connection with biological (living) systems? Obviously this must not be the case [1], a sufficiently small system consisting of an inorganic monatomic material (aluminum) shows surprisingly new effects in analogy to those which are typical for biological systems. How far has such an Al system to be reduced in order to show such features? In the micrometer realm the system behaves completely nonbiologically. However, if the size is of the order of a few nanometers, such new effects emerge. What is the adequate theoretical treatment of such and similar nanoeffects? How relevant in this context are the usual methods of solid state physics, which have been successfully applied to the bulk state? In order to be able to answer these questions more specifically, let us mention the following often observed characteristics of nanosystems. The properties of nanosystems vary strongly with temperature—much more than in the bulk of the crystal. This fact can lead to the effect that certain nanosystems (e.g., nanomachines) work properly only within a relatively

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (51–103)

Computational Atomic Nanodesign

52 small temperature window, since components (or even the whole nanosystem) lose their stability outside this temperature window. The reason for this temperature sensitivity is obvious and is explained by the following fact [2]. As the surface atoms are, in general, less bonded than bulk particles we expect that the structure and dynamics are more sensitive to variations in temperature. This is for example, the case for the mean-square displacements at the surface and also for anharmonic effects, leading to the effect of premelting; that is, the crystal is molten in the surface region already below the melting temperature. Since the relative number of surface atoms increases with decreasing size of the system, these effects are large in the case of nanosystems and, as we have already remarked, the melting temperature Tmn of the whole nanosystem is in general far below the melting temperature Tm of the bulk. Furthermore, below Tmn various specific structure transformations [3] can take place in connection with systems of nanometer size. All these effects can influence the function of nanosystems and have to be considered carefully. Also this behavior of nanosystems makes clear that the basic principles of solid state physics can hardly be extended to the realm of systems with nanometer size. For example, the concept of phonons cannot be extended to nanosystems and we have to use other methods for the determination of the particle dynamics in such systems. Only a few concepts, which work well in the description of bulk states, can be used for the characterization of nanosystems. In conclusion, the typical features of the solid in the bulk (periodic structure, harmonic behavior) can be no basis for the theoretical description of nanosystems, and the system under investigation has to be analyzed carefully with respect to the theoretical treatment. However, the principles of modern surface physics are often useful within the analysis of nanometer-scale properties. The reason is obvious: In the case of nanosystems a great fraction of atoms belong to the surface region of the system and determine essentially their behavior, and such nanosurface effects can be much larger than those of semiinfinite surface systems. For example, in the case of sufficiently small nanoclusters no particle of the system can be considered as a bulk particle since the distances of all the particles to the surface are smaller then the cutoff radius for the interaction. In this chapter we introduce methods which are adequate for the theoretical treatment of nanosystems. These methods have been discussed critically and, in some cases, proposals are made to improve them under consideration of recent research results. The material is not only an introduction to the state of the art for researchers and professors, but is also benefits advanced students. In this work we had to restrict ourselves to the study of the structure and dynamics of the atoms which form nanosystems (machines, aggregates, clusters etc.); we do not deal with nanostructured materials or with electronic properties of nanosystems (quantum dots, quantum computers, etc.). Clearly, the structure and dynamics of atoms are determined by the electrons, but only in connection with the interaction

between the particles, and these interactions are of fundamental importance for the understanding of the structural and dynamical behavior of nanosystems. In the study given here we will concentrate on the molecular dynamics method, since this method is the most powerful tool in the theoretical treatment of nanosystems; within this method the structure and dynamics of the particles can be investigated as a function of temperature. This is important since the nanometer-scale properties are in general very sensitive to small variations in temperature, as we have already outlined above. The molecular dynamics method is described in detail, and also models and methods for the description of interaction potentials which are needed in connection with molecular dynamics calculations. We also discuss the Monte Carlo method. While this tool is restricted to the study of structural properties, it has, on the other hand, the advantage that it can be used to treat quantum systems. We start our considerations with the quantum-theoretical treatment of nanosystems, that is, with the methods of quantum chemistry. Here the discussion concentrates on the Hartree–Fock method and the density-functional formalism. However, we have restricted ourselves to a few brief statements. The reason is the following: until now, these methods only allowed the determination of zerotemperature properties, and as we have already discussed, temperature effects are of considerable importance in connection with nanosystems. In particular, the nanoproperties are in general very sensitive to small variations in temperature and, therefore, it should hardly be possible to conclude from zero-temperature data (Hartree–Fock method, density-functional formalism) to non-zero-temperature properties.

2. QUANTUM-THEORETICAL TREATMENT OF NANOSYSTEMS 2.1. Approximations For sufficiently small systems, that is, systems with a particle number not larger than a few hundred, the methods of theoretical quantum chemistry are useful. The main disadvantage of these approaches is that they are only applicable for systems at zero temperature, and this is a far-reaching restriction in connection with nanosystems, since temperature effects play a central role here, at least for metallic nanosystems. Furthermore, quantum chemistry is not able to say something about the dynamics of the atoms as, for example, the phonons. The central problem of quantum chemistry is to find an approximation for the electronic Schrödinger equation. The nuclei are fixed at first; only the electrons are treated quantum mechanically (Born–Oppenheimer approximation). In a further step the displacements of the nuclei can be easily treated. The direct solution of the electronic Schrödinger equation  = E H

(1)

is not possible, either in an analytical way nor within the frame of numerical methods. The approximate solution of

Computational Atomic Nanodesign

53

(1) can only be achieved on the basis of relatively drastic simplifications for the wave function  . The loss of precision is only then acceptable when a certain inaccuracy is compensated by another; only in this way can we obtain useable results for the structure and the binding energies of the system under investigation [4].

2.1.1. Hartree–Fock Approximation For systems with more than 300 electrons only the most simple ansatz for the wave function  is applicable, and this is expressed by the Hartree–Fock approximation [5] which has the following form: HF = 1 r1 n rn

(2)

where n is the number of electrons. Each electron is modeled by a wave function i ri and includes a spin function. The brackets · · · in (2) denote antisymmetrization and consider the Pauli principle. The optimization of the wave functions i ri , which is necessary because ansatz (2) is very simple, leads within the frame of the variational procedure to the Hartree–Fock equations Fi = i i

(3)

In other words, the three-dimensional Hatree–Fock Eq. (3) replaces the 3n-dimensional Schrödinger equation (1), and this is a drastic simplification. However, the Fock operator depends on the wave functions i ri which one actually wants to know and, therefore, for the solution of (3) an iteration procedure is needed. Despite the enormous simplifications the Hartree–Fock equations are only applicable to simple cases, for example, atoms or small molecules. Therefore, it is necessary to approximate the wave functions i ri by a superposition of basis functions b r : i r =



Bi b r

(4)



This approach is called linear combination of atomic orbitals (LCAO) since for the functions b r atomic orbitals have been used in the early stage. Presently almost exclusively “contracted Gauss-type orbitals” (CGTOs) are used,

b =

  

   k l m 2 x yz d exp − r

2.1.2. Density-Functional Formalism The theoretical foundations of this method have been developed by Hohenberg and Kohn in 1964 [6]. Within the conventional treatment of the n-electron system the external potential ex r , produced by the charge of the nuclei, determines uniquely—together with the n electrons—the Hamiltonian of the system and, therefore, the properties of the ground state, at least in principle. In the work by Hohenberg and Kohn it is outlined that in the case of a nondegenerated ground state instead of n and ex r also the one-particle density n r can be used for the unique characterization of the many-electron system. The external potential ex r of the n-electron system is uniquely determined by the one-electron density n r . Also the energy E0 of the ground state is a unique functional of this density n r : E0 = E0 n r

(6)

Since the one-particle density n r determines the number of electrons (with the normalization condition n =  n r dr , it contains the information about the wave function of the ground state. Here, the external field ex r is not restricted to Coulomb potentials; ex r has merely to be local and, therefore, it can also be applied to external electrical fields. Further, a variational principle exists for the one-particle density n r : The functional of the total energy En r

becomes stationary if n r is given by the density of the ground state. Let us consider the electron density n′ r

with n′ r ≥ 0 and n = n′ r dr. Then, the energy En′ r

is always larger than the exact energy of the ground state E0  En′ r

≥ E0 , and we have En′ r

= E0 in the case of the exact ground state density n r = n′ r . If En r

can be differentiated, the Euler–Lagrange equation En r

= n r

(7)

can be deduced if the particle number n remains constant; is the chemical potential. Although the details of the density-functional formalism are not given here, it should be mentioned that Kohn and Sham proposed a method for the treatment of the kinetic energy [7]: In the treatment of the many-particle problem the kinetic energy is not easy to determine since this energy part is large in comparison to the entire energy and, therefore, the kinetic energy has to be calculated very precisely. Kohn and Sham proposed a method for that which is presently the basis for most numerical calculations in connection with the density-functional formalism.

2.1.3. Discussion (5)

which are centered at the nuclei of the system. By the use of these basis functions the Hartree–Fock method becomes an efficient procedure [4]. More details and specific techniques, respectively, are given in the solid state literature, in particular in [5].

The Hartree–Fock method begins conceptionally with the description of individual electrons interacting with the nuclei and all the other electrons of the system. In contrast to this picture within the density-functional formalism the total energy of the whole electron system is considered. There are three terms: the kinetic energy, the Coulomb energy due to the electrostatic interactions among all the charged particles, and the so-called exchange-correlation energy for which the actual expressions are unknown. Within the LDA

54 the exchange-correlation energy is taken from the manyelectron interactions in an electron system of constant density, which are known, or more specifically: Within the LDA it is assumed that the exchange-correlation energy of an inhomogeneous system is given in an infinitesimal volume dr by that of a homogeneous system. The LDA turned out to be convenient in connection with numerical calculations and is surprisingly accurate. The LDA is exact for an ideal metal where the electron density is a constant and, therefore, this approximation is less precise for systems with a varying electron density. Within the density-functional formalism instead of the electron density we have the one-electron densities, which are expressed by one-electron wave functions which are similar to those used within the Hartree–Fock approximation, that is, the wave functions i ri (Section 2.1.1). It is not clear a priori which of the two approaches, the Hartree–Fock method or the local density-functional formalism, gives a better description. For the applicability of both methods the effective interaction range of the electrons is of importance. The Hartree–Fock method works better if the interaction range is of the order of several interatomic distances. This is because the molecular orbitals, which are used in Hartree–Fock based methods for the description of correlation effects, are quite large (they are extended over several interatomic distances). On the other hand, if the range of many-body effects is short, that is, smaller than interatomic distances, then the local density-functional formalism should be applied; in such cases a description with mathematical objects like molecular orbitals is extremely slowly converging. For large molecules there is a further method which has been applied successfully [4]. This approach extends the Hartree–Fock method and is known as the MP2 approach (MP2: Möller–Plesset, second order perturbation theory). This MP2 approach can only be applied if already a good Hartree–Fock result for the system under investigation could be obtained. On the basis of this Hartree–Fock result an improved description is possible if the approximations typical for the Hartree–Fock approach are corrected by means of perturbation theory (MP2). It turned out that results based on MP2 calculations are more accurate than those obtained by the density-functional formalism [4].

2.1.4. Applications Hartree–Fock Method A lot of interesting applications exist in connection with the Hartree–Fock method. Here we would like to list some selected works that are typical for various topics in connection with this method. Specific Nanosystems Articles include: theoretical study of the catalytic properties of Pt/Fe nanoclusters [8], ion channeling on nanofilms [9], structure and thermodynamics of carbon and carbon/silicon precursors to nanostructures [10], size-dependent oxidation of hydrogenated silicon clusters [11], tomonaga–luttinger parameters for quantum wires [12], interactions in chaotic nanoparticles [13], Hartree–Fock dynamics in highly excited quantum dots [14], and Coulomb interactions in carbon nanotubes [15]. Further information is given in [16–40].

Computational Atomic Nanodesign

Surface Investigations Articles include: Hartree–Fock studies of surface properties of BaTiO3 [41], effect of basis set superposition error on the water-dimer surface [42], ab initio Hartree–Fock study on surface desorption process in tritium release [43], ab initio Hartree–Fock study of Bronsted acidity at the surface of oxides [44], and structure and bonding of bulk and surface theta-alumina from periodic Hartree–Fock calculations [45]. More information concerning surface investigations is given in [46–54]. Molecules Articles include: a comparison of finite difference and finite basis set Hartree–Fock calculations for the N2 molecule [55], exact solution of the Hartree–Fock equation for the H2 molecule in the linear-combinationof-atomic-orbitals approximation [56], from the nonplanarity of the amino group to the structural nonrigidity of the molecule [57], and coupled Hartree–Fock calculations of origin-independent magnetic properties of benzene molecule [58]. Further information is given in [59–67]. Metals Articles include: metal dissolution in aqueos electrolyte: semi-empirical Hartree–Fock and ab initio molecular dynamics calculations [68], density-functional computations of transition metal NMR chemical shifts: dramatic effects of Hartree–Fock exchange [69], electronic structure and orbital ordering in perovskite-type 3d transition metal oxides studied by Hartree–Fock band structure calculations [70], Hartree–Fock ground state of the composite fermion metal [71], and localized-orbital Hartree–Fock description of alkali-metal clusters [72]. More information about metals is given in [73–84]. Biological Systems Articles include: ab initio quantum mechanical study of hydrogen-bonded complexes of biological interest [85], combining implicit solvation models with hybrid quantum mechanical/molecular mechanical methods: a critical test with glycine [86], and an efficient coupled Hartree–Fock computational scheme for parity-violating energy differences in enantiomeric molecules [87]. Further information about biological systems is given in [88–95]. Islands Articles include: shape and stability of heteroepitaxial metallic islands: effects of the electronic configuration [96], spin depolarization in quantum dots [97], and magnetic-field dependence of the level spacing of a small electron droplet [98]. More information is given in [99–103]. Specific Application (Hartree–Fock) Ab Initio Models for ZnS Clusters The Hartree–Fock molecular orbital method has been applied in the study of ZnS clusters [104–106]. The electronic structure of ZnS is localized; this can be shown, for example, by the density matrix approximation. The influence of vacancies and cluster edges on the charge distribution have distinct damping effects on it. The influence of the ZnS vacancy on the electronic structure of the 6 × 6∗ 2 cluster is shown in Figure 1, where the cluster is labeled according to the rings and layers [104–106]. According to Figure 1 the changes are local and there is a distinct damping near the vacant site on both the sulfur and the more metallic Zn ending surfaces.

Computational Atomic Nanodesign

55 combined density functional theory and EXAFS study [130], and Structures and spectra of gold nanoclusters and quantum dot molecules [131]. Further information is given in [132–181]. Molecules Articles include: infrared spectra and density functional calculations of the SiCO4 molecule in solid argon [182], a density functional model for tuning the charge transfer between a transition metal electrode and a chemisorbed molecule via electrode potential [183], a density functional study of possible intermediates of the reaction of dioxygen molecule with non-heme iron complexes [184], and van der Waals interaction of the hydrogen molecule: an exact implicit energy density functional [185]. More information about molecules is given in [186–197].

Figure 1. Charge density isosurface of the 6 × 6∗ 2 cluster, where the ZnS monomer is removed from the middle of the zinc (left) and sulfur (right) surfaces. The isosurface is colored according to the density difference between the clean and the defect clusters. Courtesy of J. Muilu, Biomedicum, Helsinki.

Density-Functional Formalism Also, many interesting studies in connection with the density-functional formalism have been published. We have selected some of them which we think are important. The density-functional formalism has been applied to the following problems. Nanostructures Articles include: organic and inorganic nanostructures: an atomistic point of view [107], ab initio study of quantum confined unpassivated ultrathin Si films [108], surface effective-medium approach to the magnetic properties of 3d adatoms on metals [109], towards controlled production of specific carbon nanostructures—A theoretical study on structural transformations of graphitic and diamond particles [110], and semiclassical density functional theory: Strutinsky energy corrections in quantum dots [111]. More information about this topic is given in [112–123]. Nanoparticles, Nanoclusters, and Clusters, Including Cluster Methods Articles include: theoretical study of the catalytic activity of bimetallic RhCu surfaces and nanoparticles towards H2 dissociation [124], adhesion of nanoparticles to vesicles: a Brownian dynamics simulation [125], influence of quantum confinement on the electronic and magnetic properties of (Ga,Mn)As diluted magnetic semiconductors [126], first mixed valence cerium-organic trinuclear cluster as a possible molecular switch: synthesis structure and density functional calculations [127], Structures of small gold clusters cations: Ion mobility measurements versus density functional calculations [128], a theoretical study of Si4 H2 cluster with ab initio and density functional theory methods [129], new insights into the structure of supported bimetallic nanocluster catalysts prepared from carbonylated precursors: a

Biological Systems Articles include: density functional theory-based molecular dynamics of biological systems [198], a self-consistent charge density functional based tight binding scheme for large biomolecules [199], density functional theory and biomolecules: a study of glycine, alanine, and their oligopeptides [200], compared performances of the molecular orbital and density functional theories for fragments of biomolecules [201], DFT calculations on the energy thresholds of DNA damages under irradiation conditions [202], and electronic structure of wet DNA [203]. Further information about biological systems can be found in [204–229]. Droplets and Islands Articles include: heterogeneous nucleation on mesoscopic wettable particles: a hybrid density functional theory [230], electronic properties of model quantum-dot structures in zero and finite magnetic fields [231], structure and stability of superfluid He4 systems with cylindrical symmetry [232], effect of oxide vacancies on metal island nucleation [233], arsenic flux dependence of island nucleation on InAs(001) [234], and first-principles studies of kinetics in epitaxial growth of III–V semiconductors [235]. Further information is given in [236–253]. Metals Articles include: investigation of the metal binding site in methionine aminopeptidase by density functional theory [254], generalized spin orbital density functional study of multicenter metal systems [255], density functional theory study on the geometric, electronic and vibrational structures of alkali metal porphyrin complexes [256], and infrared spectra and density functional theory calculations of group 4 transition metal sulfides [257]. More information about metals is given in [258–270]. Nanotubes Articles include: first principles calculations for electronic band structure of single-walled nanotube under uniaxial strain [271], theoretical study of structuredependent Coulomb blockade in carbon nanotubes [272], theoretical tools for transport in molecular nanostructures [273], electronic structures of capped carbon nanotubes under electric fields [274], and Fullerene based devices for molecular electronics [275]. More information concerning nanotubes is given in [276–290].

Computational Atomic Nanodesign

56 Nanowires Articles include: Fe nanowires on vicinal Cu surfaces: Ab initio study [291], Pentagonal nanowires: A first principles study of the atomic and electronic structure [292], Simulation of quantum transport in nanoscale systems: Application to atomic gold and silver wires [293], Quantum transport through one-dimensional aluminum wires [294]. Further information about wires and nanowires, respectively, is given in [295–304]. Specific Application (Density-Functional Theory) 1. Electronic conductance through monatomic Na wires coupled to electrodes has been studied using the LDA and the Friedel sum rule. Conductance (G is found to exhibit even–odd behavior. When the number L of atoms in wires is odd, G has the quantized value of G0 =

e2 $

which is stable with respect to variations of the wire geometry. On the other hand, G is smaller than G0 for even L and is dependent on the wire geometry in this case. This behavior results from the charge neutrality and resonant character due to the sharp tip structure of the electrodes and may appear in other monovalent atomic wire such as gold. More details concerning this study are given in [305, 306] and in Figure 2. 2. On the basis of density-functional calculations of model systems for the active site, Rod et al. have suggested a mechanism for the ammonia synthesis catalyzed by the enzyme nitrogenase [307, 308] (see also Fig. 3). The mechanism can explain many experimental observations including the obligatory H2 release. The authors have also suggested a mechanism behind the observed CO frequency shift upon CO adsorption on the FeMo co-factor [308]. In [309] the enzyme catalyzed reaction with the reaction catalyzed by a Ru surface is compared and it is explained why the two reactions need different conditions (low temperature and pressure as opposed to high pressure

Figure 2. (a) The atomic structure for a L = 5 Na wire connected to electrodes with the inversion symmetry. (b) Contour plot of the difference of the total charge densities between the systems with and without the L = 5 wire (marked by green dots); black contours for 0.001 to 0.005 a.u. with the increment of 0.001 a.u., whereas blue and red ones for −0 0003 to 0.0003 a.u. with the increment of 0.0002 a.u. Courtesy of Heung-Sun Sim, School of Physics, Korea Institute for Advanced Study, Seoul.

Figure 3. Left panel: The MoFe protein, component of the enzyme nitrogenase, which catalyzes the ammonia synthesis from atmospheric nitrogen. The active sites, where the synthesis is going on, are depicted by atom specific coloring, while the remaining atoms are colored green. Atoms have been removed from the structure in order to make the active sites visible. Right panel: The FeMo cofactor MoFe7 S9 (homocitrate) is the active site, where N2 bins and react with protons and electrons. Courtesy of T. H. Rod, The Scripps Research Institute, Department of Molecular Biology, La Jolla.

and temperature). On the basis of that, the authors give suggestions to how the ammonia synthesis can be (electro)catalyzed at low temperature and pressure on a surface; they also discuss the properties that such a surface should have. Reference [310] is basically a summary of the above plus work by Per Siegbahn, but with emphasis on what we can learn from the cross interaction between surface science and biological science. The methods, which the authors have applied, are pretty much standard surface science density-functional theory methods, namely density functional calculations using a plane wave basis with cutoff energy of 25 Rydberg. They have used Vanderbilt ultrasoft pseudopotentials [311] and soft pseudopotentials [312] (for sulfur) to describe the core part of the atoms. The generalized gradient approximation in the form of Perdew–Wang [313] and revised Perdew–Burke–Enzerhoff [314] has been used to describe the exchange-correlation term. The FeMo co-factor is antiferromagnetic, so spin dependent exchange-correlation functionals have been used. Remarks Concerning Metallic Systems The quantum theoretical treatment of metal systems, for example, clusters, is a particular challenge. The reason is that the electronic structure of such systems is rather complex which is directly connected with the features of metals. Metal clusters obviously cannot be treated within the Hartree–Fock method [4]; accurate calculations are only available for very small clusters (up to Al6 ). However, the density-functional formalism works well and reliable for larger clusters (for example, Al146 ). More details are given in [315].

Computational Atomic Nanodesign

57 which is a unique functional of n r :

2.1.5. Density-Functional Formalism at Nonzero Temperature The Hartree–Fock method and the density-functional formalism are only able to describe systems at zero temperature. However, the structure and also the dynamics of metallic nanosystems (for example, Al clusters) are sensitive to small variations in temperature: Relatively small changes in temperature can lead to drastic effects in connection with the atomic structure and dynamics, and the properties at zero temperature can become meaningless. Therefore, the following question arises: Can the methods of theoretical quantum chemistry be extended to nonzero temperatures? In principle, this is possible within the framework of densityfunctional formalism. Since this point is important for future developments, let us discuss the situation in somewhat more detail, in particular, with respect to the changes in the formalism when we go from zero to nonzero temperatures. Within the density-functional formalism the density n r

is of relevance and its determination is in any case a quantum mechanical problem for all temperatures. Let us consider a system of N electrons of mass m and charge −e which are distributed around positive charges of magnitude  Z1 e, Z2 e at positions R1 , R2 . The Hamiltonian H is given in the second-quantized representation by  = T + U +W  H

where

2  * + + r * + r dr T = 2m

(8)

+

1  Zi Zj e 2 2 i j Ri − Rj

 = W

With (12) we have E0 n r

= where



 r n r dr + F n r

 - F n r

= - T + U

 N exp − H− k T .0 = B 

Tr exp − Hk− TN

dr  r + + r + r

(10)

(11)

(12)

For zero temperatures Kohn and co-workers [6, 7] showed that for a nondegenerate - ,  r is (to within a constant) a unique functional of n r and the correct n r minimizes the ground state energy  - E0 = - H

(18)

 is the particle number is the chemical potential, and N operator. Mermin showed [316] that in a grand canonical ensemble for a given temperature T , chemical potential , and external potential  r the quantity   /n r

= dr  r n r + Gn r

− dr n r (19) with

where  r is an external potential arising from the nuclei. The electron density n r is given in the ground state - by the expectation value n r = - + + r + r -

(16)

B

Gn r

= Tr0.0 T + U + kB T ln .0 1 

(15)

Equation (15) defines a variational principle for the ground state energy of an electron gas in an external potential  r , in which n r is the variable function. However, from this density-functional formalism we can only extract zero-temperature properties for an electron gas, which fundamentally determines the system properties; all calculations reported in literature have been done on the basis of this zero-temperature approach. It is, however, possible to extend this formalism to nonzero temperatures, and this has been done by Mermin [316]. In the case of nonzero temperatures the equilibrium electron density in a grand canonical ensemble is given by

n r = Tr .0 + + r + r

(17)

(9)

i =j

 in (8) is given by W

(14)

where .0 is the grand canonical density matrix

is the kinetic energy operator. + r and + + r are annihila describes the tion and creation operators. The operator U mutual repulsion, both among the electrons and the ions:  e2 = 1 + + r + + r ′ + r ′ + r

dr dr ′ U 2 r − r ′

E0 = E0 n r

(13)

(20)

is equal to the grand potential

    H− N / = kB T ln Tr exp − kB T

(21)

when n r , expressed by (17), is the correct equilibrium density. In particular, Merwin showed that the correct density minimizes (19) over all density functions that can be associated with an external potential  r . Since the properties of systems with surface (for example nanosystems) are very sensitive to variations in temperature, zero-temperature approaches, such as the Kohn– Hohenberg–Sham formalism, are not appropriate for the treatment of such systems (in particular metallic nanosystems). It is, in our opinion, in many cases an insufficient approximation to use zero-temperature methods for

Computational Atomic Nanodesign

58 the determination of the electronic properties, which determine chiefly the entire properties of the system. Such an analysis is only correct for zero temperatures but has often been used for the determination of the structure and dynamics for nonzero temperatures, in particular in connection with molecular dynamics calculations (see Sections 5.3.4 and 5.3.5, embedded-atom method and quantum molecular dynamics introduced by Car and Parrinello). Such calculations can be misleading since nonzero temperature data are coupled with zero-temperature properties. Therefore, instead of the Kohn–Hohenberg–Sham approach Mermin’s formalism, briefly discussed above, should be used in such calculations. But to the authors knowledge there exists no analysis on the basis on Mermin’s non-zero-temperature formalism.

3. BASIC PROPERTIES AT THE NANOLEVEL Specific material properties of nanosystems can be essentially different from the corresponding properties of macroscopic systems (the solid in the bulk). This has been demonstrate in [317] in connection with the thermal stability and the melting temperature of specific nanostructures. An essential reason for this tendency is the fact that a great fraction of the particles (atoms, molecules) of such small systems belong to the surface region and the surface particles are less bonded than the particles in the bulk, leading to relatively strong anharmonicities even at low temperatures, that is, far below the melting temperature. Even the melting process takes place far below the bulk melting temperature or is simply not defined as has been demonstrated in connection with the nanosystem discussed in [317]. The thermal behavior of such systems is a complex function of the particle number and the outer shape of the systems. This must have consequences for the theoretical description of the material properties for systems of nanometer size. Let us discuss this point in more detail.

3.1. Relevance of the Theoretical Bulk Methods for the Description of Nanosystems In modern materials research, solid-state physics becomes more and more relevant, in particular, the microscopic structure and dynamics. For sufficiently large systems, the standard model of solid-state physics is in most cases adequate for the description of material properties. The standard model of solid-state physics is based on the ordered structure and on the assumption that the vibrational amplitudes of the atoms are sufficiently small, so that it can work within the harmonic approximation. In other words, the dynamics is expressed by phonons. For example, the specific heat at constant volume is expressed in terms of phonons by the expression [318, 319] cV = kB

 62 exp 62

2 p q exp 6 − 1

(22)

with 6=

7p q

kB T

(23)

where 7p q are the phonon frequencies, q is the wave vector, and p labels the phonon branches. Electronic properties are of course also essential. For example, for the understanding of superconductivity, the electron–phonon interaction is important. In conclusion, within conventional materials research, the basis for the microscopic description of the material properties is the ordered structure and the harmonic approximation; the properties of a solid are given in terms of phonons 7p q , where the phonon frequencies 7p q are dependent on the crystal structure and the electronic properties. That is the situation in connection with macroscopic systems. But this can no longer be the basis for the description of the properties of nanosystems. As was outlined in [317], nanosystems do not behave in such a simple way: They are in most cases disordered and the dynamics cannot be approximated by phonons because the harmonic approximation is not applicable. Therefore, at the nanolevel materials research gets a new dimension. Even the melting temperature is no longer a fixed material property, as is demonstrated in connection with the nanostructure discussed in [320]. Nanosystems are already disordered at relatively low temperatures and the harmonic approximation breaks down completely; that is, even at relatively low temperatures, phonons are no longer definable. In such cases the anharmonicities cannot be considered as small perturbations to the harmonic case. In other words, the standard model of solid-state physics breaks down at the nanolevel in many cases, and we have to introduce other methods and quantities for the description of material properties. This is the case for any anharmonic and disordered system (for example, liquids in the bulk) and, for simplicity, we want to explain the principal points by means of a bulk liquid. In the case of liquids, usual crystallography and phonons are not relevant but the structure must be described by statistical mechanics, that is, in terms of correlation functions, for example, the pair correlation function g r , which can be measured; r is the relative distance between the atoms and molecules, respectively. In the case of metals, the interaction is given by the pair potential  r , where r is again the relative distance between two particles. In conclusion, instead of the phonons 7p q , correlation functions like g r and the interaction potential  r

become the relevant quantities: 7p q → g r  r

For example, the isothermal compressibility 9T is defined by the expression [320] 1 1 =− 9T V



:p :V



(24) T

Computational Atomic Nanodesign

59

and we need an expression for the pressure p (V is the volume); that is, we need the “equation of state,” which can be formulated in terms of g r and  r as follows: p = .kB T −

.2  : r

g r dr r 6 :r

(25)

If we apply Eq. (25) to Eq. (24), the isothermal compressibility 9T can be expressed in terms of g r and  r : g r  r → 9T In the case of nanosystems, the anharmonicities are very strong and the thermal expansion becomes a relevant quantity. Moreover, a great fraction of the atoms belong to the surface region and it turns out that the thermal expansion at the surface is distinctly larger than in the bulk. Therefore, the theoretical picture must be able to describe the thermal expansion. The thermal expansion coefficient 6p can be expressed by the isothermal compressibility 9T and the thermal pressure coefficient V : 6p = 9T V

(26)

In order to get the thermal expansion coefficient 6p , we have to calculate both the quantities 9T and V . We already did that for 9T , and for the description of   :p (27) V = :T V we need again the pressure p, that is, the equation of state, which is given for the bulk by Eq. (25). In other words, the thermal expansion coefficient 6p can also be expressed in terms of g r and  r , g r  r → 6p and this is the case for all material properties (in some cases, higher order correlation functions are needed). It should be mentioned that the thermal expansion is zero for the harmonic case; that is, within the phonon picture the thermal expansion cannot be described. In conclusion, in the case of disordered systems with strong anharmonicities (e.g., liquids in the bulk and nanosystems), phonons are not suitable for the adequate description of such systems but must be expressed in terms of g r

and  r . In the case of nanosystems, a great fraction of particles belong to surface regions, and the quantities of g r and  r are not simply dependent on the relative distance r of the particles but are also dependent on the distance of the particles from the surface. Because there is in general more than one surface in the vicinity of a particle the situation can be very complex. Even for systems with only one surface—for example, a semi-infinite liquid—the situation is already rather complex. In this case, the surface (defining the x, y plane) separates the liquid from the vapor phase, and the density . varies with respect to the coordinate z (the coordinate perpendicular to the surface), so that we have . = . z .

In the liquid–vapor transition zone, the pressure p cannot be described by Eq. (25); that is, p is no longer a constant and independent of the coordinates but becomes an anisotropic quantity [321]: At any point z in the transition zone, we have a normal pressure component p1 and a tangential component p2 . While p1 and p2 are different from each other in the liquid–vapor transition zone both quantities are of course identical in the isotropic bulk phases. With r12 = x12 y12 z12 , the statistical mechanical expressions for p1 and p2 take the form [321, 322] 1 z2 . z 2 . 2 r z *  r 12 dr (28) 2 r 2 1 x p2 = . z kB T − . z 2 . 2 r z *  r 12 dr (29) 2 r p1 = . z kB T −

where we assumed for simplicity that the pair potential at the surface is the same as in the bulk. In Eqs. (28) and (29) the distribution function . 2 r z is used instead of the correlation function g r z . The pressure components p1 and p2 can be used for the determination of the surface tension  which is expressed by  p1 − p2 z

dz (30)  =  T = −

With Eqs. (28) and (29), the statistical mechanical expression for the surface tension is expressed by  T =

x2 − z212 1   2

dr dz . r z *  r 12 2 − r

(31)

As outlined in [321], the surface tension and its temperature dependence are of particular interest in the description of thermodynamic properties. In the case of nanosystems, the situation concerning the statistical mechanical description is more complex than for semi-infinite systems. This is because various surfaces, which are most often close together, have to be considered. Therefore, the statistical mechanical expressions [in analogy to Eqs. (28), (29) and (31)] will hardly be relevant in connection with theoretical investigations. A certain potential for applications should be given in the case of disordered nanofilms that are positioned on surfaces. The more direct way is to perform molecular dynamics calculations (details of this computational method are given in the next section). In such calculations the interaction potential, in particular the pair potential  r , is of fundamental relevance and, therefore, in Section 5 we will give a brief overview of this topic.

3.2. Wear at the Nanolevel Within the frame of Hamiltonian systems, such as molecular dynamics systems, friction in the macroscopic sense is not defined. At the microscopic, molecular dynamics level, the forces are formulated as quantities which are dependent on the structural configuration (particle positions) but not on the particle velocities. Therefore, a force which is proportional to the velocity cannot be introduced at the microscopic level and, therefore, a friction constant in the macroscopic sense is not definable at the microscopic

Computational Atomic Nanodesign

60 level of modern materials research. In the case of nonHamiltonian systems friction can be studied by nonequilibrium molecular dynamics; this topic is briefly discussed in Section 4.2.4. At the nanolevel, wear is described by specific complex processes. Let us discuss this point by means of an example: Figure 4 shows a molecular dynamics model for a wheel which is pressed on a surface. The wheel rotates with 1012 revolutions in a second. The wheel has a temperature of 300 K and it has a diameter of approximately 10 nm. If the wheel is pressed on a surface, wear effects emerge and, according to the magnitude of the force vertically applied, the wheel can even be destroyed, as can be observed in Figure 4. In summary, friction (wear) at the microscopic level is a complex process and cannot be characterized by only one constant as is the case of macroscopic friction. In our example (Fig. 4), wear is dependent on the specific structure of the surface and also on the shape of the wheel.

3.3. Conclusions Theoretical materials research at the nanolevel not only needs solid-state physics but due to the strong anharmonicities and the strongly disturbed structure of many nanosystems (even far below the melting temperature of the corresponding bulk state) the basics of statistical mechanics and the theory of liquids are very often more appropriate as the basis for the theoretical description of such systems. In such cases the standard model of solid-state physics [ordered structure and harmonic approximation, here summarized by the symbol 7p q ] is not suitable, and we come to the description in terms of correlation functions and interaction potentials that we have briefly characterized by g r and  r . Within this statistical mechanical description time correlations are of course included, for example, the velocity autocorrelation function whose Fourier transform has to be considered as the generalized phonon density of states.

However, due to the strong anharmonicities, the dynamics of such systems is characterized by a broad range of different dynamical states including small local vibrations and complex diffusion processes. This situation indicates that the development of a suitable standard model for the theoretical description of nanosystems will hardly be possible; even in the case of liquids in the bulk a standard model could not be found up to now. Therefore, the most important tool for the investigation of nanosystems is the molecular dynamics method since anharmonicities are treated within this method without approximation, and this is important because the typical anharmonicities in connection with nanosystems cannot be considered as small perturbations to the harmonic approximation. No other microscopic method in condensed matter physics allows one to treat anharmonicities without approximation; this is only possible in connection with phenomenological models. Thus, the molecular dynamics methods should be considered as the standard method for the theoretical description of nanosystems. On the basis of these considerations we come to the following rough classification scheme: 7p q

(standard model for the solid) ↓

g r  r

(statistical mechanics)

(32)



molecular dynamics (standard method for nanosystems) In other words, with decreasing particle number disorder effects and anharmonicities increase and this has an influence on the theoretical treatment. In Section 4.2.3 more details concerning the molecular dynamics method will be given.

4. MODELING OF NANOSYSTEMS 4.1. The Hamiltonian 4.1.1. General Formulation The quantum mechanical modeling of a system with N particles of masses m leads to the Hamiltonian = H

Figure 4. A molecular dynamics model for a wheel consisting of aluminum atoms; it rotates with 1012 rev/s and its diameter is approximately 10 nm. The molecular dynamics method is discussed in more detail in Section 4.

N  i=1





 N  2 2 *i + Vi ri + Vik ri rk

2mi i k=1

(33)

i =k

Here Vi ri is an externally applied potential in which the ith particle is located, where ri = xi yi zi is the position of the ith particle; Vik ri rk denotes the interaction potential between the two particles i and k. To analyze or to describe the characteristics of the system one has to solve the according many-particle Schrödinger equation + = E+ H

(34)

Computational Atomic Nanodesign

61

where E is the total energy of the N -particle system. The wave function + depends on the 3N coordinates (configuration space) of all particles: + = + x1 y1 z1 xN yN zN

(35)

If we consider nanosystems, most often external potentials are not present and the particles involved are atoms which in turn have to be divided into nuclei and electrons. In this case the interaction potential in Eq. (33) is given by the Coulomb potential Vik ri rk =

Z i Zk e 2 ri − rk

(36)

where Z is the according particle charge number including the sign of the charge. With a closer look at this many-particle problem it becomes clear that an exact quantum mechanical solution can probably never be achieved. Here is an example: A relatively small nanocluster of only 100 argon atoms consists of 100 nuclei and 1800 electrons, which is a total of 1900 particles. In this case the configuration space consists of 5700 dimensions. The key point with numerical solutions of the Schrödinger equation is the spatial integration. With the assumption that a division of each dimension into 100 steps is sufficient for an accurate calculation we would have to compute the summation of 1011400 volume elements. Needless to say, this is not possible without further, most intensive simplifications and approximations.

4.1.2. Approximations The treatment of many-particle systems in condensed matter physics is usually based on the fact that core electrons do not influence the properties of such many-particle systems and, therefore, the system can be divided into ions and valence electrons. A further approximation is to treat the ions and electrons separately (Born–Oppenheimer approximation) which leads to two Schrödinger equations—one for the electrons, where the ion positions appear as parameters only, and another for the nuclei, where the electron energy Eel acts as an effective potential—so that the Hamiltonian for the ions takes the form

with

= H

NI   i=1





2 2 * + Vi ri + V r1 rNI

2mi i

V r1 rNI = VI r1 rNI + Eel r1 rNI

(37)

Under the assumption that the change of the electronic arrangement (core electrons) around each ion or atom is sufficiently small the potential energy V may be expanded: V r1 rNI =

NI NI 1  1  ij +  + ··· 2 i j=1 6 i j k=1 ijk i =j

(39)

i =j =k

Here the terms on the right of Eq. (39) represent pair, triplet, and many-body contributions of the ionic or atomic interactions and r1 rN are the positions of the N ions or atoms. If the polarization of the core electrons is negligibly small, then—compared with the pair terms—the triplet and higher terms diminish rapidly in significance. Then the obvious final step of approximation is to neglect them entirely. This is called the pair potential approximation, V r1 rNI =

NI NI 1  1  ij =  ri − rj

2 i j=1 2 i j=1 i =j

(40)

i =j

with ri − rj = rij . Referring to our example with the 100 argon atoms, with Eq. (40) the problem has been reduced to a 9900-fold sum of values from one pair potential function  with only one dimension, which is the distance of two particles. This simplification expands the calculability of the many-particle problem with today’s computer power up to millions of particles—at least under certain conditions. On the other hand, such a far-reaching, if not to say brute simplification, has a strong influence on the applicability as can be easily imagined [323]. In almost all cases the ions or atoms behave classically and, furthermore, only the pair interaction is effective. Then, instead of the Hamiltonian (39) we get the classical Hamilton function in the pair potential approximation. For monatomic systems it reads H=

NI NI  1  pi2 +  rij

2m 2 i j=1 i=1

(41)

i =j

If three or higher body forces are effective, instead of the pair potential approximation for the potential energy the expansion (39) has to be used. Pair potentials and manybody forces will be discussed in more detail below.

4.2. Description of Nanosystems (38)

where NI is the number of ions. Equation (38) means that the dynamics of the ions take place within an effective potential V which is formed by the potential energy of the direct ion– ion interaction VI and the entire electronic energy Eel , where Eel depends on the coordinates of the ions and therefore plays the role of an indirect ion–ion interaction. But it must be emphasized that an unique splitting into ions and valence electrons is not always possible.

4.2.1. Simple Models How can we determine the properties of a many-particle system on the basis of a given Hamiltonian? In the conventional treatment of this problem we need simple models. Quite generally, simple models can be obtained by controlled approximation from the general cases (37) and (41). In connection with many-particle systems the general mathematical formulation of the problem is in most cases so complex that simplifying models have to be chosen but often cannot be obtained by controlled simplifying steps from the

Computational Atomic Nanodesign

62 general cases (37) and (41), respectively. It is therefore a rule to introduce simple models just for convenience. The “simple model” of solid state physics is the crystalline solid in the harmonic approximation (Section 3.1). On the basis of this model one is able to determine successfully the properties of a lot of materials. However, there are also a lot of cases where this simple model is not applicable, even when it is extended by specific assumptions. For example, the silver subsystem of the solid electrolyte 6-AgI is highly disordered and shows strongly anharmonic behavior [324–326]; a simple model for 6-AgI and similar materials is not available. Also in the case of liquids and gases “simple models” have not been found. Even in the case of gases with low densities we cannot simply restrict ourselves to the first terms in the virial expansion in the calculation of the pressure. This is because the expansion obviously converges slowly and, therefore, even in the case of low density one has to consider more than the first two terms [327]. The virial coefficients can be expressed in terms of the pair potential  r [see Eq. (40)], but the expressions are getting complicated for the higher order virial coefficients and in practical calculations only the first terms are accessible. The virial expansion would define a “simple model” if for a broad class of gases a restriction on the first two terms would be realistic. In summary, only for a specific class of many-particle systems could a “simple model” be found: It is the crystalline solid in the harmonic approximation, but in most cases anharnonicities cannot be considered as small perturbations to the harmonic approximation (see also the discussion in Section 3.1). This is also the case for typical nanosystems since a great fraction of atoms are more or less close to the surface region of the system, and surface particles behave strongly anharmonically. In small nanosystems (for example, clusters with a few hundred of atoms) even the innermost particles cannot be treated as bulk particles since the cutoff radius for the interaction potential is larger than the distance to the surface. The atoms at surfaces are less bonded than in the bulk and, therefore, the mean-square amplitudes of the atoms are significantly larger than at the surface leading to relatively strong anharmonicities. This phenomenon can be observed even at low temperatures—low in comparison to the melting temperature of the bulk system.

4.2.2. Dynamical-Matrix Solutions Surfaces not only give rise to quantitative effects but can also be the source for new qualitative phenomena. For example, the phonons, which describe the dynamics of the system, are expressed by the dynamical-matrix solutions, and these solutions are obviously not complete for systems with surfaces. In order to show that in more detail let us consider a monatomic crystal with a free surface and let M be the mass of the atoms of the crystal. In this semi-infinite system there is only a two-dimensional invariance of the force constants which implies that the normal-mode solutions of the equation of motion u¨ 6 l = −

 1 ?6> l l′ u> l′

M ′ l >

6 > = x y z

(42)

is a superposition of two-dimensional Bloch functions 

l u6 l12 m = (43) u6 j m q exp i q·r012 −7j q t n1 n2 j

where l = l1 l2 m . Here u6 is the displacement from the equilibrium positions in the 6-direction, l12 = l1 l2 specifies the points in the layers parallel to the surface, and m labels the planes of the crystal which are parallel to the surface starting with m = 1 for the outermost layer. The frequencies 7j q are given by the solutions of the dynamical matrix, where j is the branch index. Further, n1 , n2 , denotes summation over the wave vectors q = n1 n2 ; the allowed values of n1 and n2 are determined by the periodic boundary conditions: ni = 0 ±1 ±Ni /2 i = 1 2 where N1 is the particle number in the x-direction and N2 in the y-direction of layer m. Now we consider the center-of mass velocity of layer m, and we want to distinguish between the bulk and surface contributions. Summation of u6=z l12 m over all the particles in layer m and the use of well-known identities yields  u˙ z l12 m = b z m + s z m

(44) z m = l12

where

a z m = −i



7k q uz k m q exp0−i7k q t1

n1 n 2 k

× N1 N2 n1 0 n2 0

a = b s

(45)

which leads to a z m = −i



7k 0 uz k m 0 exp0−i7k 0 t1

k

a = b s

(46)

b z m and s z m are the contributions of the bulk modes b and the surface modes s to the center-of-mass motion. b z m is by definition not localized at the surface and, therefore, it is independent of m. From Eq. (46) follows that only those surface modes contribute to s z m for which 7 = 0 in the limit of q → 0; that is, only optical surface modes contribute to s z m (the acoustic surface phonons only give rise to a uniform translational motion at q → 0). From the analysis of the motion of the atoms relative to each other follows that the excitation described by the amplitude A m of s z m cannot be an optical surface mode. Moreover, it has been discussed in literature (see, for example, [328]) that in the long-wave limit monatomic systems with only one particle in the three-dimensional unit cell admit only acoustic surface waves which do not contribute to s z m . In other words, the dynamical-matrix solutions would lead to s z m = 0

(47)

and the center-of-mass motion is entirely given by b z m

and must be independent of m. However, molecular dynamics calculations (see Section 4.2.3) show that the amplitude A m of the center-of-mass motion is strongly dependent on m [323]. In particular, it is shown in [323] that this

Computational Atomic Nanodesign

63

excitation is localized at the surface: Its amplitude A m

decreases with increasing m, and its penetration depth into the crystal is only a few interatomic distances. All this is in contrast to Eq. (47). In other words, the center-ofmass motion at the surface, described by s z m , cannot be described by the dynamical-matrix solutions, and it is shown in [323] that this effect is large; the dynamical-matrix solutions are obviously not complete for systems with surfaces, and in connection with nanosystems a relatively great fraction of atoms belong to surface regions.

4.2.3. Molecular Dynamics The relatively strong anharmonicities in nanosystems are not negligible even at low temperatures. Furthermore, we have discussed in the last section in connection with the centerof-mass motion that surfaces can be the source for new qualitative phenomena. Since the dynamical-matrix solutions are obviously not complete the following questions arise: How can the dynamics of nanosystems be described? What method or basic model is adequate? Since only for the ideal solid does such a basic model exist [other simplifying principles at the microscopic level could not be found for many-particle systems (Section 4.2.1)], we have to describe such systems on the basis of the most general formulation. In the case of classical systems this can be done on the basis of Eq. (41) if the pair potential approximation can be applied and if many-body interactions can be neglected [Eq. (39)]. Such a general description can be achieved within the framework of molecular dynamics calculations which have to be performed for the characterization of nanosystems. We already came in Section 3.1 to this conclusion [scheme (32)]. In the following we would like to give more details concerning the molecular dynamics method. Basic Information Only on the basis of the general formulation the relatively strong anharmonicities can be treated without approximation. In other words, the general description of the properties of classical many-particle systems (for example, nanosystems) can be done by means of molecular dynamics calculations without the use of “simple models” (Section 4.2.1) or other simplifying assumptions (Fig. 5). As already mentioned several times, such “simple models” and simplifying assumptions, respectively, are not known at the microscopic level and can only be introduced in a phenomenological or empirical way. Within the framework of molecular dynamics Hamilton’s equations dH dH p˙ xi = − p˙ zN = − dx1 dzN dH dH x˙1 = z˙ N = dpxi dpzNi

Figure 5. If a “simple model” is available, the properties of many-particle systems—including those of nanometer size—can be determined on the basis of such a model; in most cases additional assumptions are necessary. The advantage of molecular dynamics is that the description can be done without simple models and other additional assumptions. This is important since in most cases such systems are so complex that simplifying models have often been chosen for convenience and could not always be obtained by controlled simplifying steps from the general case (41). For nanosystems a simple model could not be found up to now. Therefore, the molecular dynamics methods is of particular importance in the nanometer-scale realm.

N = NI atoms: q t1 p t1

q t2 p t2



(49)

q ti p ti

where q ti = x1 y1 z1 xN yN zN

  p ti = px1 py1 pz1 pxN pyN pzN

(50)

In other words, as solutions of the classical equations of motion we obtain the coordinates and momenta for all particles of the system as a function of time t. The time step in the iteration process is Et = ti+1 − ti . With k iteration steps the information about the system consisting of N particles is given in the time interval F = k · Et. Equation (49) contains the total information of the many-particle system. On the basis of this information the properties of the system, in particular experimental data, can be determined—at least in principle. In the following we will give some remarks about the determination of typical functions on the basis of information (49). Correlation Functions In the analysis of many-particle systems correlation functions

(48)

are solved by iteration with the help of a high speed computer, and we obtain the following information for the

a t ′ b t ′′

(51)

a t ′ = a q t ′ p t ′

(52)

of two quantities a t ′ and b t ′′ are of particular interest, where b t ′′ = b q t ′′ p t ′′

Computational Atomic Nanodesign

64 The angular brackets · · · in (51) denote a thermodynamic average, which will be discussed in connection with molecular dynamics calculations in the next section. The time evolution of the system is given in statistical mechanics by the  operator S t :  ′ − t ′′ a q t ′′ p t ′′

a q t ′ p t ′

= S t

(53)

 = exp i S t

Lt

(54)

 can be written in terms of the Liouville operawhere S t

 tor L, with

 3N   :H : :H : − :qi :pi :pi :qi i=1

(55)

a t ′ b t ′′ = a 0 b t ′′ − t ′

(56)

 L=i

where qi and pi denote all the 3N coordinates and momenta, respectively. In general it is difficult to determine the time evolution of many-particle systems on the basis of Eq. (54). However, on the basis of the molecular dynamics information (49) it is straightforward to determine correlation functions of type (51). In connection with time correlation functions of type (51) it is important to mention that they are not dependent on the time origin; that is, in Eq. (51) t ′′ − t ′ are relevant but not t ′ and t ′′ :

Let us briefly discuss an example: With a 0 = v 0 and b t = v t we get the velocity autocorrelation function - t = v 0 · v t

(57)

where v t is the velocity at time t for one atom of the ensemble. With - t = - −t the Fourier transform of - t

is given by f 7 =

2 v 0 2 $





0

- t cos 7t dt

(58)

where f 7 is normalized to unity: 

 0

f 7 d7 = 1

(59)

In the case of the harmonic solid f 7 is just the frequency spectrum of the normal modes, that is, the phonons. The frequency spectrum f 7 defined by (57) is quite general and is applicable to systems with strong anharmonicities like nanosystems. f 7 describes the complete dynamics of many-particle systems; that is, all kinds of excitation are included—even diffusion processes. The diffusion constant is directly expressed by f 7 [329]: D=

$ kB T f 7 = 0

2 m

(60)

In particular, using f 7 in the description of the dynamics no problems appear in connection with systems with surfaces where the dynamical matrix solutions are obviously not complete as we have demonstrated in Section 4.2.2. How can we calculate the velocity autocorrelation function - t and other properties from the basic molecular dynamics information (49)? In order to answer this question we have to outline how thermodynamic averages · · · can be treated within the framework of molecular dynamics, which we have introduced in the context with Eq. (51). In the next section we will give some principal remarks. Average Values Imagine a space of 6N dimensions whose points are determined by 3N coordinates q ti = x1 y1 z1 xN yN zN and 3N momenta p ti = px1 py1 pz1 pxN pyN pzN . This space is the so-called phase space where each point at time t corresponds to a mechanical state of the system. The evolution with time of the system is completely determined by Hamilton’s equations (40) and is represented by a trajectory in phase space. The trajectory passes through the element dq dp at point (q,p) of the phase space and it can pass it several times. This process defines a “cloud” of phase points indicating how often the elements of the phase space have passed through by the trajectory. In other words, instead of the trajectory we have now a “cloud” of phase points. The “cloud” is a great number of systems of the same nature, but differing in the configurations and momenta which they have at a given instant. In summary, instead of considering a single dynamic system we consider a collection of systems, all corresponding to the same Hamilton function. This collection of systems is the so-called statistical ensemble (see also, for example, [330–337]). The introduction of the statistical ensemble is very useful in regard to the relationship between dynamics and thermodynamics. The statistical ensemble can be described by a density function . q1 q3N p1 p3N

(61)

in phase space. The number of points in the statistical ensemble is arbitrary and, therefore, . q p has to be normalized:  . q p t dq dp = 1 (62)

The quantity

. q p t dq dp

(63)

can be considered as the probability of finding at time t a system of the ensemble in the element dq dp at the point (q,p) of the phase space. The introduction of the density (61) is meaningful only if the density takes an asymptotic value for long times t: lim . q p t = . q p

t→

(64)

where . q p is the density function of the statistical ensemble. When the trajectory q t , p t is able to produce, over

Computational Atomic Nanodesign

65

a long period of time, the function . q p (ergodic hypothesis) then we have equivalence between the time average and the average over the statistical ensemble of a function f q t p t

: 

f q p . q p dq dp  . q p dq dp 1x f t dt f = lim t→ F 0

f =

(65) (66)

Different expressions exist for the density function . q p . These expressions depend on the thermodynamic environment: (1) microcanonical ensemble (an isolated system, N , V , and E are given, where E is the energy and V the volume); (2) canonical ensemble (a closed isothermal system, N , V , and T are given, where T is the temperature); (3) grand canonical ensemble (V , T , and given, where is the chemical potential). In molecular dynamics calculations in most cases N , V , and E are given and, therefore, such models represent microcanonical ensembles. In connection with nanosystems the following discussion is relevant: Within the thermodynamic limit N → V →

(67)

N = const V the microcanonical, the canonical, and the grand canonical ensemble are equivalent; that is, within the thermodynamic limit we obtain in all three cases the same value for the statistical average of f of f q p . This is not fulfilled in the case of nanosystems since the properties of nanosystems are in general dependent on the particle number N . For nanosystems two points are important: (1) What can we say about the dependence of the particle number N ? (2) What does a long period of time [F →  in Eq. (66)] mean? Discussion Concerning Point (1) The properties of nanosystems are in general dependent on the particle number. This is the reason we investigate nanosystems. In order to study this point in more detail one has to study the statistical average g of a suitable quantity g q p as a function of N and if there is g N ′ = g N ′′ = g N ′′′

(68)

N ′ < N ′′ < N ′′′

(69)

with

we may consider the system under investigation for N ≥ N ′ no longer as a nanosystem. On the other hand, if N ′ is the critical particle number we may consider the system for N < N ′ as a nanosystem.

Discussion Concerning Point (2) With Eq. (64) and using the well-known expression for the velocity distribution    3/2 m mv2 . q p = . v = 4$ (70) v2 exp − 2$kB T 2kB T we obtain for the mean-square velocity 3/2      mv2 m v2 = 4$ dv v4 exp − 2$kB T 2kB T 0 =3

kB T m

(71)

On the other hand let us consider a single particle of the N -particle system and let vi t be the velocity of the ith particle at time t where the index i has been chosen arbitrarily. If we define a function v2 F by F vi2 t dt (72) vi2 F = 0

we expect [according to (65) and (66)] that lim vi2 F = v2

F→

(73)

That is, there is equivalence between the average over the time and the average over the statistical ensemble [Eq. (71)]. In other words, we expect that after a sufficiently large time F the velocity vi t of a single atom of the N -particle system, which can be arbitrarily chosen, has passed through all the states given by Maxwell’s distribution (70). In connection with molecular dynamics calculation the following question is of particular interest: How large is the time F in order to fulfill (73) in a good approximation? Specific calculations showed that F is relatively small; to fulfill Eq. (73) one needs not more than 103 iteration steps (i.e., F is of the order of 10−11 s if the time step is 10−14 s). Clearly, within molecular dynamics calculations averages will not be calculated on the basis of a statistical ensemble (65) or with Eq. (66). The calculated information expressed by (49) already represents the data in thermal equilibrium. Therefore, it is straightforward to determine averages on the basis of (49). For the mean-square velocity [see Eq. (71)] the molecular dynamical average is v2 =

NL N  2 1 1  v L

NL N j=1 i=i i j

(74)

The average is formed over all N particles and various times Lj . Time Evolution of Molecular Dynamics Systems From the solutions of Hamilton’s equations (48) we obtain the coordinates and momenta (velocities) of all N particles as a function of time [see Eq. (49)]; this is the total microscopical information of the system under investigation. However, the solutions of Hamilton’s equations require initial values for the coordinates and the velocities of the N particles. In the case of gases and liquids the initial values for the coordinates can be distributed randomly with the appropriate density. In connection with crystals (with and without surfaces, and nanosystems) the particles will be situated

Computational Atomic Nanodesign

66 within the array so that the perfect lattice structure appropriate to the system under investigation is generated. Without external forces acting on the system, the directions of the initial values for the velocities should be distributed randomly so that the momentum of the whole system is approximately zero at the initial time t0 , and it must remain zero for all times t (conservation of momentum). In thermal equilibrium the magnitudes of the particle velocities are distributed according to Maxwell’s distribution. However, it is more convenient to choose for all particles the same magnitude of the velocities, and this means that the system is initially not in thermal equilibrium. With the help of the function

The temperature of molecular dynamics systems is dependent on time t and is defined by the relation

2 1 N vi t 2 6 t = N i=1 N 1 2 2 i=1 vi t

N

As in the case of 6 t also the temperature T t fluctuates in finite systems around the mean temperature (see Fig. 7): 1 L T t dt (81) T = lim L→ L − L1 L1

(75)

where vi t , i = 1 N are the velocities obtained from the molecular dynamics calculations, we can study at which point in time Maxwell’s distribution (70) is reached. In the case of (71) the function 6 t in (75) takes the value of 6 t =

5 3

(76)

for all times t. However, if all particles have the same initial magnitude of the velocities we get for 6 t at time t0 6 t = 1

(77)

It can be demonstrated by realistic molecular dynamics calculations that thermal equilibrium (6 t = 5/3 is reached after a few hundred time steps if we start from 6 t0 = 1. Due to the finite number of particles 6(t fluctuates around its equilibrium value of 5/3 (see also Fig. 6); these fluctuations are getting small with increasing particle number N and are physically realistic. Clearly, in the case of macroscopic systems N →  the fluctuations are zero.

Figure 6. Schematic representation of 6 t . The system is initially t0 = 0 not in thermal equilibrium (6 = 5/3). After a few hundred molecular dynamics time steps thermal equilibrium (Maxwell’s distribution) is reached. Due to the finite number of particles the system fluctuates around 5/3.

3 1 m v t 2 = kB T t

2 2

(78)

With v t 2 =

N 1  v t 2 N i=1 i

(79)

we get for the temperature as a function of time T t =

N m  v t 2 3NkB i=1 i

(80)

L1 is the time where the system has reached equilibrium, and L is the time of investigation (i.e., for which Hamilton’s equations have been solved after equilibrium is reached). As in the case of 6 t , the temperature fluctuations (see Fig. 7) are physically real. From these fluctuations we can extract the specific heat at constant volume (per particle)   1 :E (82) cV = N :T V It is straightforward to show that in the case of a microcanonical ensemble the specific heat is expressed by 338 339   3 3N 2 2 −1 (83) T − T

1− cV = 2 2T2 where T and T2 are obtained by averaging over a sufficiently large time interval. For example, T2 is given by 1 L T2 = lim T t 2 dt (84) L→ L − L1 L1

Figure 7. The temperature as a function of time (schematic representation). T is the mean temperature; L1 is the time where the system has reached equilibrium. Due to the finite number of particle the temperature fluctuates around its mean value T .

Computational Atomic Nanodesign

In other words, the temperature fluctuations are a direct measure of the specific heat, indicating that these fluctuations are not an artificial element of the method.

4.2.4. Nonequilibrium Molecular Dynamics A few remarks should be given in connection with nonequilibrium molecular dynamics which has been developed in addition to equilibrium molecular dynamics (Section 4.2.3) and which has been known since the early 1970s [340–342]. This method was introduced in order to compute efficiently transport coefficients. To establish the nonequilibrium situation of interest, an external force is applied to the system. Then the response of the system to these forces is determined from the simulation. This method has been used for the calculation of diffusion coefficients, the shear and bulk viscosity, and thermal conductivity [343]. Further basic literature concerning nonequilibrium molecular dynamics is given by [344–349]. Recent interesting results can also be found in [350–355]. Non-Hamiltonian systems can be studied by means of nonequilibrium molecular dynamics [356], for example, dissipative systems, that is, systems which involve friction in one of its various forms. In such cases, particle trajectories are calculated from the equations of motion which are, however, not consistent with any Hamilton function.

5. INTERACTION POTENTIALS Clearly, in order to be able to perform molecular dynamics calculations not only are the initial values for the coordinates and momenta (velocities) for all the N particles of the system needed as input (Section 4.2.3), but in particular the interaction potential between the particles, which reflects the specific characteristics of the system under investigation. For a reliable description of material properties precise knowledge of the interaction potential is required. We will outline below that surface properties in general, as well as in connection with nanosystems, are particularly sensitive to potential variations. Both the repulsive and the attractive parts of the potential have to be determined carefully. However, the determination of accurate pair potentials for systems of interest (e.g., metals and materials with covalent bonding) is rather difficult even for the bulk, and the nanoproblem, where a great fraction of atoms belong to the surface region, is much harder because of the change in electronic states and other relevant properties near the surfaces of the nanosystem. In this section we would like to discuss methods which— from the authors point of view—are relevant for the determination of interaction potentials. In literature a lot of theoretical studies can be found—in particular, in connection with molecular dynamics calculations—in which the potentials are used in a rather uncritical way: The potentials have not been tested carefully, in some cases even not at all. Serious research in nanoscience requires just the opposite: Since the properties of nanosystems are particularly sensitive to potential variations, the potentials have to be determined thoroughly for such systems.

67 5.1. Types of Interactions in Condensed Matter Physics Basically, in condensed matter physics we have to distinguish between four binding types (see, for example, [357]): (1) ionic interactions (4–14 eV), (2) metallic bonds (0.7–6 eV), (3) van der Waals interactions (0.02–0.3 eV), and (4) covalent bonds (1–10 eV), where the numbers denote the typical binding energies. Also the hydrogen bridge bond (0.1–0.5 eV) is a characteristic binding type but will not be discussed in the following.

5.1.1. Ionic Binding The ionic interaction is characterized by different atoms (for example, Na and Cl) which have exchanged electrons, so that the many-particle systems consist of positive and negative charged ions. The interaction law as a function of distance r between the ions is simply described by Coulomb’s law if the distances are larger than r+ + r− , where r+ and r− are the radii of the positively and negatively charged particles. For distances r < r+ + r− the interaction is repulsive due to the overlap of the electron cores.

5.1.2. Metallic Binding In the case of certain atoms (for example, Al) the electrons of the outer shell are only weakly bonded. If a many-particle system is formed with such atoms the weakly bonded electrons leave the according atoms and move through the whole system. These are the conduction electrons. In other words, in metals we have the following situation: There are positively charged ions which move through the sea of conduction electrons. In metals the ion–ion interaction is not simply given by the Coulomb potential since the ions are screened by the conduction electrons and that makes its determination complicated. Such a modeling of the ion–ion potential can be performed, for example, within the frame of pseudopotential theory; in Section 5.3.1 we give some details concerning this method.

5.1.3. Van der Waals Interaction The motion of the electrons around the atomic nucleus means that there is a certain probability that the atoms have at instant t an electrical dipole moment which, however, becomes zero when it is averaged over the time. In a manyparticle system these momentary dipole moments interact with each other leading to an attractive potential between the atoms of the system. Typical van der Waals systems are noble gases. Clearly, repulsive core effects have to be considered also in the description of the interaction potential at small distances.

5.1.4. Covalent Bonds In the case of covalent binding the interaction between particles is given by the fact that part of their electrons belong to several particles at the same instant. The probability of finding an electron, which is responsible for binding, is relatively large in the environment between the particles leading to the interaction between them. In most cases this process is based on the formation of spin saturated electron

Computational Atomic Nanodesign

68 pairs, where each atom contributes an electron so that the electron shells take a noble gas configuration. Examples for substances with covalent bonds are carbon, amorphous semiconductors, hydrogen molecules, etc. Remarks Concerning Phenomenological Potentials In connection with realistic calculations phenomenological potential functions are used very often, which are most often not strictly constructed with respect to the basic interaction types discussed above. As a typical example for a phenomenological potential let us briefly discuss the potential of Morse [358] which has been extensively used in the study of lattice dynamics [359], the defect structure in metals [360–368], the inert gases in metals [369, 370], the equation of state [371, 372], elastic properties of metals [373], the interaction between gas atoms and crystal surfaces [374], and in studies of other specific problems [375–381]. In order to calculate the energy levels for diatomic molecules, Morse required some conditions for the interatomic potential of the atoms of the molecule [323] which describes the spectroscopy of the molecules. Morse chose the potential function  r = 60 exp−26 r − r0

− 260 exp−6 r − r0

(85)

where r0 is the intermolecular separation. The solution of the radial part of Schrödinger’s equation using this potential yielded the correct energy level representation. In conclusion, the Morse potential has not been constructed on the basic interaction types discussed above but is based on the spectroscopy of molecules. Nevertheless, the potential has been used extensively in the study of various many-particle properties, but there is no physical justification for that. There are many other phenomenological approaches for the interaction (for example, the Buckingham potential [382–385]), but we do not want to discuss these functions here because in most cases there is no or almost no physical background for their introduction; these potentials have often been chosen for purely pragmatic reasons and, therefore, they have to be considered only as fitting functions. Since there is a strong potential sensitivity in connection with the properties of systems with surfaces (for example, nanosystems), we have to construct the potentials very carefully. In the next sections we would like to discuss some relevant theoretical methods for the construction of model potentials.

Let us demonstrate that by means of krypton [386]. In connection with nanosystems surface properties are of particular interest and, therefore, let us study the sensitivity of surface properties on the potential function. As is well known, krypton is a noble gas, and in contrast to metals, the potential functions of noble gas systems (solid, liquids, nanosystems, etc.) do not depend on density and temperature, respectively, and they are the same at the surface as in the bulk of the crystal. At free metal surfaces the local background may be changed from its average bulk value and produces concomitant changes in the potential functions. Such difficulties do not arise at noble gas surfaces and, therefore, in the study of the potential sensitivity we want to restrict our study to noble gas (krypton) surfaces. For krypton a reliable potential, the so-called Barker potential, is available [387]; its form is rather complex and will not be outlined here. The Barker potential is more realistic than the famous Lennard-Jones potential  12  6  M M  r = 4 − r r

(86)

because it undoubtedly correlates accurately a much wider range of experimental data. Although the shapes of the Lennard-Jones potential and the Barker potential are similar (see Fig. 8) the results for the relaxation are qualitatively different from each other: The Barker potential leads to a contraction and, on the other hand, the Lennard-Jones potential leads to an expansion, as has been demonstrated by molecular dynamics calculations for the Kr (111) surface and 539 atoms [388]. The contraction is confirmed by low energy electron diffraction (LEED) data [388]. Figure 9 shows that the molecular dynamics model calculations agree well with the experimental data. The interlayer spacing d⊥ at the surface is always smaller than the corresponding value for the bulk within the temperature range studied. As already mentioned above, the lattice contraction resulting from the model and the experiments is in contradiction to the results of model calculations with the LennardJones potential (leading to an expansion), indicating that investigations—for example, in connection with dynamical

5.2. Sensitivity of System Properties against Potential Variations Before we discuss specific model potentials let us first give some remarks about the sensitivity of the system properties on the interaction between the particles which form the system. It turned out that the properties of many-particle systems are very sensitive to small variations in the interaction potentials. Therefore, in the description of system properties on the basis of the particle interaction the potentials have to be modeled very carefully. What does “sensitive to small variations in the interaction potentials” mean?

Figure 8. Pair potentials for krypton: (a) Barker potential, (b) LennardJones potential.

Computational Atomic Nanodesign

69 (Axilrod–Teller potential [390]) 3 i j k = 

Figure 9. Temperature dependence of the interlayer spacing d⊥ measured by LEED experiments ♦ for electron energies E = 53 eV (a) and E = 125 eV (b) together with the results obtained by molecular dynamics calculations ( ) [in (a) for the outermost interlaying spacing and in (b) averaged for the outermost two layer spacings]. The solid lines show the interlayer spacings for the bulk as determined by X-ray experiments. The error bars in the experimental data are mainly due to the error for the effective inner potential. The error bars with respect to the molecular dynamics results are due to the statistical fluctuations.

processes in nanosystems—on the basis of the LennardJones potential can be seriously in error. It should be emphasized here that the Lennard-Jones potential is the most used interaction function in literature, and sometimes it is even used for the description of metals [389], but there is no physical basis for that since the interaction type (van der Waals forces) described by the LennardJones potential has nothing to do with the interaction type dominant in metals. In conclusion, the properties of many-particle systems (in particular those with surfaces like nanosystems) are obviously very sensitive to small variations in the interaction between the particles. We have shown that the relatively small quantitative differences between the Lennard-Jones potential and the Barker potential (Fig. 8) give rise not only to quantitative effects but also to qualitative differences. This example shows that one has to be very careful in the selection of interatomic potentials, especially when modeling nanosystems, since a great fraction of particles belong to the surface region. With respect to the interaction problem we have still to mention that the Lennard-Jones potential as well as the Barker potential are descriptions within the pair potential approximation (Section 4.1.2). However, in principle manybody forces [three-body interactions etc.; see Eq. (39)] might also play a role. What can we say about the influence of three-body interactions? The calculations in connection with Figure 9 have been done for the noble gas krypton which is a typical van der Waals system. For such systems a three-body potential exists

1 + 3 cos L1 cos L2 cos L3 rij3 ril3 rjl3

(87)

where  is a constant; rij , ril , rjl and L1 , L2 , L3 are the sides and angles of the triangle formed by the particles i, j and l. 3 i j l can be repulsive and attractive, which depends on the shape of the triangle formed by the three atoms i, j, and l. Are those Axilrod–Teller interactions relevant in connection with the results given in Figure 9? We cannot exclude that the three-body potential (87) together with the Lennard-Jones potential may lead to good results for the relaxation at the surface. However, it is shown in [391] that the influence of 3 i j l on structural data is relatively small and, therefore, we cannot expect that the Lennard-Jones potential together with 3 i j l gives good results for the relaxation and other quantities of relevance. On the other hand, the Barker potential is expressed by a lot of parameters, and these parameters were fitted by a lot of experimental data; this fact can be expressed formally by  =  r a b

(88)

where a, b, etc. are the fit parameters. In other words, such kinds of potentials should not be considered as pure pair potentials in the sense of Eq. (39) in Section 4.1.2 but rather play the role of effective interaction potentials  r a b = eff r a b which in particular consider approximately many-body forces. Therefore, we may express Eq. (39) in Section 4.1.2 by V r1 rN =

N N 1  1  + ··· ij + 2 i j 6 i j k ijk i =j



i =j =k

N 1  r a b

2 i j eff ij

(89)

i =j

That is, the effective potential simulates approximately many-body forces. This is due to the fact that the potential parameters a, b, etc. are fitted to experimental data which may be more or less sensitive to many-body forces.

5.3. On the Determination of Pair Potentials in Metals 5.3.1. Remarks Concerning Pseudopotential Theory As outlined in Section 5.1 metals consist of ions of positive charge and electrons. In the case of the ionic interaction the potential between the particles is simple and is given by the Coulomb potential. In metals the situation is more complicated since the ions are surrounded of conduction electrons. The ions are screened by the electrons; in particular, there are quantum-mechanical correlation and exchange effects among the electrons. All these facts have to be considered in the construction of ion–ion potentials  r

in metals, where r is the distance between the two ions of the system. Within the framework of pseudopotential theory

Computational Atomic Nanodesign

70 (see, for example, [392]) all these effects (screening, correlation, and exchange effects) can be adequately considered. The pseudopotential method is often used for the determination of the interaction between the ions in metals. Within the frame of this theory the problem can be treated on various levels, and a lot of pseudopotential models exist in literature. Let us briefly discuss here a typical example for a monatomic system. This example shows how all the previously mentioned effects (screening, correlation, and exchange effects) can be treated mathematically and how they are interrelated. As already outlined, the metal system has to be considered as a conglomerate of ions which are embedded in an uniform compensating background of negative charge (conduction electrons). Within the framework of pseudopotential theory the ion–ion potential  r can be divided into two parts [392, 393],

to zero in Eq. (94),  q is equal to the static Hartree dielectric function. It should be mentioned that there exist various models for the dielectric function  q in literature [392], and the form of  q essentially depends on the material under investigation. The Fourier transform w0 R of the matrix element k + q w0 k can be considered as the unscreened potential 0 R

between the ion and the conduction electron at distance R. 0 R must be corrected by an orthogonalization potential orth R in the core region due to the orthogonalization of the plane waves to the core states [392],

(90)

0 R in Eq. (96) is expressed by the Coulomb interaction between the conduction electron and the excessive point charge Ze of the ion:

 r = d r + ind r

where d r is the so-called direct interaction between the ions, and ind r is the ion–electron–ion interaction, which has the meaning of an indirect interaction. In most cases d r is approximated by the Coulomb potential d r =

Z ∗2 e2 r

(91)

of the point charges Z ∗ e [392]. The deviations of the effective valency Z ∗ from the valency Z are in most cases less than 10% [393]. The difference Z ∗ –Z can be thought of as resulting from the orthogonalization of plane waves to the core states which gives an extra charge on the ion. The local approximation of the indirect ion–ion interaction ind r [see Eq. (90)] can be expressed by [392] /0   sin qr 2 F q

q dq $2 0 qr

(92)

 q − 1 /0 q 2 k + q k 2 8$e2  q 1 − C q

(93)

ind r =

Following Sham [394] and Heine and Abarenkov [395] for the function F q the following ansatz can be chosen: F q = −

w0 R = 0 R + orth R

(96)

where w0 R is related to k + q w0 k by k + q w0 k =

4$   sin qR 2 w0 R

R dR /0 0 qR

0 R = −

Ze2 R

(97)

(98)

The function orth R in Eq. (96) has to be modeled. For example, within the often used Ashcroft model [396] we have orth R =

Ze2 1 − L R − rc

R

(99)

with  =0 L R − rc

=1

R ≤ rc R > rc

(100)

where rc is the core radius. Within the Ashcroft model the Coulomb potential is canceled within the core region. A typical feature of the ion–ion potential  r in metals is that, due to the conduction electrons, the long-range part oscillates. This is schematically shown in Figure 10 in comparison to a Lennard-Jones-type potential which

with    me2 1−x2  1+x   q = 1+1−C q

1− ln 2$kF  2 x2 2x 1−x 

(94)

and

C q =

2q 2

q2 + kF2 + ks2

(95)

where /0 is the specific volume, kF is the Fermi wave number vector, x = q/2kF , ks = 2kF /$a0 , and a0 is the Bohr radius. The matrix element k + q w0 k is the unscreened pseudopotential form factor, which is independent of k in the local approximation. With C q exchange and correlation effects among the electrons are taken into account approximately in the dielectric function  q . If C q is taken equal

Figure 10. The schematic representation of different potential types: (a) the ion–ion interaction  r for a metal; (b) the typical atom–atom potential in a noble gas.

Computational Atomic Nanodesign

71

describes noble gases, that is, systems without conduction electrons. Equations (90)–(100) represent the basic model, but it turned out [397] that there can obviously be additional effects which are due to the cores of the ions. These are van der Waals-type interactions. Without going into detail the analysis of this problem [397] leads to additional interaction terms with respect to the direct interaction d r

between the ions [see Eq. (91)] and the unscreened interaction 0 R between the conduction electron and the ion [see Eq. (96)], 6 6 Z ∗2 e2 − 61 − 42 r r r Ze2 6 − 4 0 R = − R R

d r =

(101) (102)

where the r −6 term in (101) represents a dipole–dipole interaction, and the r −4 and R−4 terms in (101) and (102) reflect monopole–dipole potentials. The constants 6, 61 , and 62 can be estimated quantum mechanically [397]. Pseudopotentials have often been applied to various problems. Some typical articles are: Avoided crossings in the interaction of a Xe Rydberg atom with a metal surface [398], The theoretical spin-orbit structure of the RbCs molecule [399], ab initio study of the adsorption of In on the Si(001)(2 × 2) surface [400], ab initio study of iron and Cr/Fe001 [401], and Dilute Bose gas in a quasi-two-dimensional trap [402]. Further information is given in [403–415].

5.3.2. Model-Independent Potentials The interaction models which are based on pseudopotential theory work well in many cases [392, 397], provided the various terms, which compose the model, have been carefully prepared. Nevertheless, such models are complicated and contain a lot of assumptions and simplifications which are often based on more or less uncontrolled calculation steps. Therefore, it would be desirable to find methods which allow one to determine model-independent interaction potentials. Such a way is still less developed but should attract more attention in the future since the properties of systems are in general very sensitive to small variations in the interaction potential (Section 5.2); this is in particular the case in connection with nanosystems where a great fraction of the particles belong to the surface of the system. However, model-independent potentials can already be determined for liquids. The according method could in principle also be applied to nanofilms. Because of its principal meaning let us briefly discuss this method. The method [416] is based on experimental data for the liquid structure and allows one to determine the interaction without the use of any parameter in a self-consistent way. Since the method is model-independent, it is applicable for all interaction types. The liquid structure is characterized by the pair correlation function g r , which can be determined experimentally. g r is an essential quantity because important thermodynamic functions (pressure, energy, etc.) are only dependent on g r —provided we can work within the pair potential approximation.

Let us start from the expansion for g r in powers of the particle number density n (see, for example, [417, 418]),    r

g r = exp −  r

(103) kB T with

 r = 1 + na r + n2 b r + · · ·

(104)

where kB is Boltzman’s constant and T is the absolute temperature of the system. The coefficients of each power of n are cluster integrals (more details concerning Eq. (104) are given in [417]). The detailed form of  r will only have a small influence on  r because  r plays the role of an integral quantity in the determination of  r . This means that different pair potentials can lead to the same function  r —at least in principle. In other words, if we know a potential 1 r =  r which satisfies the condition (105)

1 r   r

we can calculate the unknown pair potential  r by   g r

 r  1 r − kB T ln (106) g1 r

where

   r

1 r

g1 r = exp − 1 kB T

(107)

and g r in (106) is the experimentally determined g r . Since Eq. (107) is in general an expansion with slow convergence, g1 r should be calculated from the known potential 1 r by molecular dynamics and not by (107) on the basis of (104). It is difficult to fulfill Eq. (105) in the first step and, therefore, instead of (106) we get the following iteration scheme:   g r

2 r = 1 r − kB T ln g1 r

  g r

3 r = 2 r − kB T ln g2 r



i r = i−1 r − kB T ln i r = i−1 r

with

(108) 

g r

gi−1 r



=  r

gi−1 r = g r

(109)

The unknown pair correlation functions gk r , k = 1 i − 1, can be calculated using k r , k = 1 , i − 1, by molecular dynamics. After the (i − 1)th iteration the molecular dynamical calculated pair correlation function gi−1 r agrees with the experimental g r [Eq. (109)] within the limit of accuracy dictated by the molecular dynamics calculation. For the applicability of the self-consistent method convergence of the scheme (108) is necessary. This depends on

Computational Atomic Nanodesign

72 the potential 1 r which we have to know in the first iteration step. In all calculations (see, for example, [416]) the approximation for low densities [ r = 1, Eq. (103)] 1 r = −kB T ln g r

(110)

worked well, where g r is again the experimental pair correlation function. The pair potential  r is sensitive to small variations in g r which means that g r has to be measured very accurately. This has been done for liquid gallium [419]. In [416] the self-consistent method has been applied to liquid gallium; the details are discussed there. The self-consistent method could also be applied to nanosystems. For example, it should be possible to determine the pair potential for the particles of nanostructures on surfaces. On the other hand, other model-independent methods should be developed which correlate the potential with suitable measurable quantities, that is, methods which are not based on the experimentally determined pair correlation function g r but on other relevant experimental input data leading to new schemes.

The pseudopotential theory as discussed in Section 5.3.1 leads to more or less phenomenological potentials with parameters which can be fitted to experimental data, in particular to data for systems with surfaces, for example, nanosystems. However, the surface properties can be considered from the beginning if we work more microscopically within the framework of first principles calculations. Let us briefly discuss this point. The electron density in metals must drop from a characteristic interior average value within the system to zero outside the surface. The ion–ion potential in metals is dependent on the electron density, and the charge redistribution at the surface has the effect that the potential at the surface is different from that in the bulk of metals. How can we determine the pair potential at metal surfaces, that is, in connection with nanosystems? Let us give an example. For sufficiently low temperatures the pair interactions outside and inside the metal surface have been calculated microscopically in [420] on the basis of the static electronic response function. The basic points of the approach are the following: An ensemble of ions with the positions Ri , i = 1 N , are surrounded by a neutralizing electron charge of density n r . An infinitesimal displacement of ion k with the position Rk introduces a perturbating potential Vk r − Rk and an energy change which is given by  1   dr dr ′ Vi r − Ri

2 Ri Rj

× 9 r r ′ Vj r − Rj

is introduced as a variable. According to Kohn, Hohenberg, and Sham [6, 7] n(r) is given in the ground state by n r = with



N 

-∗i r -i r

(111)

where 9 r r ′ is the static electronic response function appropriate to the electronic density n r . n(r) can be determined by the density functional formalism (see also Sections 2.1.2 and 2.1.5). How is the function 9 r r ′ correlated to the quantities defined by the density functional theory? As we have already outlined in Section 2.1.5, within

(113)

i=1

n r dr = N

The wave functions -i r satisfy 

2 2 * + Veff r -i r = i -i r

− 2m

(114)

(115)

where Veff r = V r + VH n r

+ Vxc n r

5.3.3. Microscopic Considerations

EE = −

this approach instead of the complicated wave function + r1 rN for the N -particles the electron density   ∗ dr1 ···drN N i=1  r −ri + r1 rN + r1 rN

n r = ∗ + r1 rN + r1 rN

(112)

(116)

with V (r) being an arbitrary external potential. VH n r

and Vxc n r

are the usual Hartree and exchangecorrelation potentials which are functionals of n(r). The static susceptibility 9 r r ′ in (111) is a ground state response and, therefore, it is a functional of n(r) and can be described by -i r . On the basis of (111) the ion–ion interaction outside and inside a metal jellium surface has been calculated [420]. In order to check the reliability of such potentials it would be most helpful to use them in molecular dynamics calculations for the description of nanosystems and compare the results carefully with experimental data. In this way, valuable insight into the validity of such potentials can be achieved. Another approach, extensively used in connection with molecular dynamics, is given by the so-called embedded-atom method; in the next section we discuss the basic principles of this method.

5.3.4. Embedded-Atom Method An alternative to the pair potential description has been proposed by Daw and Baskes [421, 422]; as in Section 5.3.3 also within this approach density-functional ideas are used. This approach, the so-called embedded-atom method, has also been used in connection with molecular dynamics for the description of various complex physical situations. First we give a brief description of the method and after that we quote some applications. The basic idea behind the embedded-atom method is the following (see, for example, [423]): The density-functional formalism allows one to formulate the total electronic energy for an arbitrary arrangement of nuclei as a unique functional of the total electron density. The method starts with the approximation that the total electron density in a metal can be expressed by the linear superposition of contributions from the individual atoms. Then, the electron density around each atom can be described as a sum of the

Computational Atomic Nanodesign

73

density contributed by all the surrounding atoms plus the electronic density of the atom in question. The first contribution is assumed to be a slowly varying function of position. If the problem is simplified by the assumption that this background electron density is a constant, the energy of the atom in question is given by the energy associated with the electron density of the atom plus the constant background density. On the basis of this ideas the total energy of a system with N atoms is given by   N N   1   u rij  (117) Etot = Fi .¯ i + 2 j=1 i=1 j =i

where Fi .¯ i is the energy required to embed atom i into the background electron density .¯ i at side i, and u rij is the core–core pair interaction potential between atom i and j separated by the distance rij . As already remarked, the host electron density .¯ i is approximated by a linear superposition of the spherically averaged electron densities of the ni atoms neighboring atom i, .¯ i =

N 

. rij

(118)

j=1 j =i

where . rij is the electron density of atom j at a distance rij from the nucleus of atom i. If the atomic density . r

and the pair interaction u r are both known, the embedding energy F can be uniquely defined by matching a certain equation for the cohesive energy of the metal as a function ¯ to experof the lattice constant. This can be used to fit F .

imental results. The embedded-atom method has been applied with success to several problems: for face-centered cubic (fcc) metals and their alloys [423], phonon spectra in metals [424], and the structure of liquid metals [425]. It has also successfully been used for the determination of surface energies and geometries including reconstruction phenomena [426]. Other relevant topics have been studied on the basis of the embedded-atom method. In this connection let us quote the following selected studies: Numerical evaluation of the exact phase diagram of an empirical Hamiltonian: embeddedatom model for the Au–Ni system [427], Development of modified embedded-atom potentials for the Cu–Ag system [428], Atomistic process on hydrogen embrittlement of a single crystal of nickel by the embedded-atom method [429], Kinetic pathways from embedded-atom-method potentials: Influence of the activation barriers [430], and Structural and thermodynamic properties of liquid transition metals with different embedded-atom method models [431]. More information concerning the embedded-atom method is given in [432–444] and in Section 6.1.

5.3.5. Quantum Molecular Dynamics An interesting procedure has been developed by Car and Parrinello (see [445] and, for example, [446–448]) which directly combines quantum theory with classical mechanics: The equations of the density-functional formalism [6, 7] are

solved simultaneously to the classical equations of motion, so that a time-consuming iterative self-consistent procedure is not needed. This method, designated as quantum molecular dynamics, is equivalent to the self-consistent solution of the equations of the density-functional theory. Within quantum molecular dynamics the classical Lagrange equations of the atomic positions Rj and velocities R˙ j are extended by a fictitious dynamics of the Kohn–Sham wave functions -i [6, 7] and their time derivations -˙ i : 2 1     dr -˙ i r  + mj R˙ 2j − E0- r 1 0Ri 1 L= 2 j i    Qij dr -∗i r -i r − ij + (119) i j

The first two terms describe the kinetic energy of the nuclei and that of the orbital motion with the fictitious mass . The third term contains the functional of the energy as an element of the density functional formalism. The fourth term of Eq. (119) guarantees orthonormality of the wave functions, where the quantities Qij are the Lagrange multipliers. Using this Lagrange equation we obtain the equation of motion for the nuclei ¨ j = − :E mj R (120) :Rj as well as the “equation of motion” for the electronic system: -¨ i r t = −

 :E Qij -j r t

+ :-i r t

j

(121)

Equation (121) enables one to determine the wave functions -i r t as a function of time with a simultaneous motion of the atoms. It must be emphasized that this equation describes the fictitious dynamics of the electronic system and does not reflect the motion of the electrons in the field of the nuclei. The essential difference between quantum molecular dynamics and classical molecular dynamics is that within the classical approach the forces are calculated by more or less empirically determined potentials (see Sections 5.3.1 and 5.3.2), whereas in the quantum case the forces are directly calculated from the quantum-mechanically determined potential ER1 R2 of the system. More details concerning the principles of quantum molecular dynamics are given, for example, in [447, 448]. For a system of about 100 atoms such calculations (accurately performed) are about 1000 times slower than force field calculations and the time scale is reduced considerably. Therefore, the Car–Parrinello method in general is reduced to the picosecond time range. Selected Studies Articles include: Wannier function analysis for understanding disordered structures generated using Car–Parrinello molecular dynamics [449], Hybrid Car– Parrinello/molecular dynamics modelling of transition metal complexes: Structure, dynamics and reactivity [450], Car– Parrinello molecular dynamics simulation of liquid water [451], A Hamiltonian electrostatic coupling scheme for hybrid Car–Parrinello molecular dynamics simulations [452], and Nonadiabatic Car–Parrinello molecular dynamics [453]. More details are given in [454–467] and in Section 6.1.

Computational Atomic Nanodesign

74 Final Remarks Molecular dynamics is in any case a nonzero-temperature method. On the other hand, within quantum molecular dynamics the density-functional formalism is used which is a zero-temperature approach and, therefore, in the case of quantum molecular dynamics nonzero-temperature data are coupled to the zero-temperature properties of the electronic part of the system. Since systems with surfaces (for example nanosystems) are in general very sensitive to small variations in temperature (in particular in connection with metals) such a treatment can lead to misleading and less relevant results, respectively, because non-zero-temperature data are mixed with zero-temperature properties. Therefore, instead of the density-functional formalism given within the Kohn– Hohenberg–Sham approach (Section 2.1.2) Mermin’s formalism, briefly discussed in Section 2.1.5, should be used within quantum molecular dynamics calculations. To the authors’ knowledge no calculations are reported yet on the basis on Mermin’s non-zero-temperature formalism.

5.4. Covalent Binding Many-body potentials have to be used mainly for the description of interactions within covalent bonded systems and materials, respectively, and for phenomena at higher energy levels. In metals many-body forces are less relevant because the interactions due to the conduction electrons (Section 5.3.1) are dominant. Allinger developed widely used models (MM2, MM3) for a broad range of organic structures [468–471] which are based on the separation of the total interactions between the particles, that are significant for a bond, into a sum of single potentials with respect to bond stretching, bond angle bending, and bond torsion. Within the MM2 model nonbonded interactions are described by the well-known Buckingham (exp-6) potential (Section 5.1). For the development of the interaction of such bond models a lot of a priori knowledge is necessary to classify all the different bond types and to treat the special cases. The MM2 and MM3 models are established to characterize within molecular mechanics the minimum energy configuration of structures, their stiffness, bearing properties, and the like [471], and they have become standards in literature. The term “molecular mechanics” means the following: “Many of the properties of molecular systems are determined by the molecular potential energy function. Molecular mechanics models approximate this function as a sum of 2-atom, 3-atom, and 4-atom terms, each determined by the geometries and bonds of the component atoms. The 2-atom and 3-atom terms describing bonded interactions roughly correspond to linear springs” [471]. Molecular mechanics studies are given, for example, in [471]. Because of the basic meaning of bond stretching, bond angle bending, and bond torsion let us briefly discuss often used models for these interaction types [471]: Bonds tend toward an equilibrium length r0 and, therefore, they resist stretching and compression. Bond stretching is treated with cubic pair potentials; that is, within the MM2

model anharmonicities are taken into account. The potential energy of bond stretching is given by the expression 1 k r − r0 2 01 − kcubic r − r0 1 (122) 2 s where r is the actual bond length, ks is the stretching stiffness, and kcubic is the parameter for describing anharmonicities. Bond angle bending is modeled in the form of sextic three-body potentials (MM2 model) depending on the angle between two bonds to a shared atom. The potential energy vL for this model is given by vs =

1 k L − L0 2 01 + ksextic L − L0 4 1 (123) 2 L where L0 is the equilibrium bond angle, L is actual bond angle, kL is the angular spring constant, and ksextic is the sextic bending constant. Bond torsion follows an expression depending on the torsion angle between two bonds within the same plane and a third one, thus acting as four-body potential; this potential describes the variation in energy associated with rotation about a bond and is expressed by vL =

v7 =

1 0V 1 + cos 7 + V2 1 − cos 27

2 1 + V3 1 + cos 37 1

(124)

Parameters for all the models described by (122), (123), and (124) are quoted for some common bond types in [471].

5.4.1. Models for Many-Body Potentials Due to their complexity and difficult derivation there are basically just a few three-body potentials available which are well suited for molecular dynamics calculations and Monte Carlo studies. A comparative study and discussion of some of these empirical potentials, which have been applied mainly to carbon and silicon and other semiconductor materials, is given in [472] and the references therein. The potentials discussed here are those of Pearson, Takai, Halicioglu, and Tiller (PTHT) [473], Stillinger and Weber (SW) [474], Biswas and Hamann (BH) [475], Tersoff (T2 and T3) [476, 477], and Dodson (DOD) [478]. These potentials constitute a good representative sample among the existing potentials (Brenner’s potential will be discussed seperately); they differ in their degree of physical level, mathematical form, and range of interaction. In [472] the potentials have been tested with respect to clusters (for example, Si2 –Si6 , bulk states, and surface properties. Similarities and differences between these potentials are discussed by a systematic comparison, although a theoretical justification for the potential functions can hardly be given. They have to be considered more or less as fitting functions, since they have mostly been chosen for purely pragmatic reasons. The potentials given by PTHT, BH, and SW can be classified as cluster potentials, the others as cluster functionals. The SW, DOD, T2, and T3 models only consider first-nearest neighbors (by the bond bending term secondneighbor interactions are implicitly considered). Interactions up to the third and seventh shell, respectively, are included in the potentials of BH and PTHT.

Computational Atomic Nanodesign

75

Cluster Potentials The cluster potentials describe bonding by means of classical two- and three-body interactions. The potential energy V can be written as

V =

′ 1 ′ V r + V3 rij rik rjk

2 i j 2 ij i j k

(125)

where all summations are distinct which is indicated by the primes. In connection with the PTHT model the potential V3 i j k is symmetric with respect to an interchange of i, j, k and we have to use i < j < k in the triple sum. The BH and SW models are only symmetric in j and k and the condition j < k has to be used. The two-body potential is generally given by V2 r = fc r A1 -1 r − A2 -2 r

(126)

where fc is a cutoff function and -s , s = 1 2, decays monotonically with r. The forms of the three-body interactions differ from each other and will be given for each potential discussed next. The PTHT Potential Within this model the functions which define V2 and V3 are given by  =1 fc r

=0

V3 rij rik Li =

2 

Zs +s rij +s rik gs Li

(128)

s=1

  +s r = exp −6s r 2 fc r

gs Li = cos Li − cos -0 s+1 The BH model has been used for the study of bulk and surface effects, for example, in connection with microclusters and bulk point defects [475]. More details are given in [472], in particular with respect to the determination of the parameters. The SW Potential The radial function -s r is identical with that of the PTHT potential. The other expressions are given by    = exp if r ≤ Rc fc r

r − Rc  =0 otherwise V3 rij rik rjk = Z+ rij + rik g Li

+ r = fc r

(129)

6

g Li = cos Li − cos L0 2

if r < Rc otherwise -s r = r −Rs

function fc a Fermi-like expression; the three-body potential function is separable. In particular, we have  

  r − M −1  = 1 + exp if r ≤ Rc fc r

 = 0 otherwise   -s r = exp −Rs r 2 s = 1 2

s = 1 2

V3 rij rik rjk = Z+ rij + rik + rjk g Li Lj Lk

(127)

fc r

r3 g Li Lj Lk = 1 + 3 cos Li cos Lj cos Lk + r =

where Rc is the cutoff radius and L1 , L2 , and L3 are the angles of the triangle formed by the particles i, j, and k. Within the PTHT model the two-body interaction V2 is the Lennard-Jones potential (Section 5.2) since the exponents of -s r were chosen to be R1 = 12 and R2 = 6. V3 is the Axilrod–Teller potential (Section 5.2). The use of these potentials for the description of V2 and V3 is more than doubtful since both potentials are appropriate for noble gases but not to describe bonding in covalent system; there is simply no physical justification for that. Thus, these potentials must be viewed as fitting functions. The PTHT model has been extensively used in the description of various materials, with and without surfaces, and it has been extended to binary and ternary systems (e.g., GaAs); more details are given in [472]. The BH Potential Within this model for the radial functions -s r Gaussians have been used, and for the cutoff

fc not only acts as a cutoff function but also defines the attractive branch of V2 (since R2 = 0) and the radial components of V3 and, therefore, the model is very sensitive to small variations in Rc . The SW potential is mostly used in the description of systems with covalent bonds, for example, in connection with clusters, lattice dynamics, bulk point defects, etc. The potential has been applied to Si, Ge, sulfur, and fluorine and to the Si–F system [472]. Cluster Functionals Of particular interest are the cluster functionals of Tersoff type, where the bonding is described by means of pairwise interactions, but many-body interactions are effectively included with respect to the attractive term of the pair interaction which is dependent on the local environment. The potential energy takes the form V =

1 ′ f r 0A1 -1 rij − A2 -2 rij p Sij 1 2 i j c ij

(130)

where Sij is the effective coordination number and determines p which is a measure of the bond order, where we have to consider that p has in general the property p Sij = p Sji

Sij is given by Sij =



k =i j

V3 rij rik Li

(131)

Computational Atomic Nanodesign

76 with V3 rij rik Li = + rij rik g Li

(132)

Rewriting (130) the two-body energy term can be extracted and we get V =

1 ′ 1 ′ V r + A - r f r

2 i j 2 ij 2 i j 2 2 ij c ij

× 01 − p Sij 1

(133)

It can be shown by simple manipulations that the second term of (133) effectively includes many-body effects [472]. The DOD Potential Dodson chose the following form for the model functions:     1 $ r − Rc

  = 1 − cos     2 if Rc − < r < Rc fc r

  if r ≤ Rc − = 1   = 0 otherwise -s r = exp −Rs r

s = 1 2  n p Sij = exp −Sij   -2 rik fc rik 6 + rij rik = -2 rij fc rij

g Li =

(134)

This potential has been used for studying surface reconstructions and low-energy beam deposition of Si. It should be mentioned that the DOD potential is similar to the first Tersoff potential [479]. The Tersoff Potential (T2 and T3 ) Here two different parametrizations are used for the same potential, leading to the T2 and T3 potentials. The expressions for fc and -s are identical with those used within the DOD potential. The other functions have been chosen as follows: p Sij = 1 + Sijn −1/2n

2 2 g Li = > + 2 − 2   + cos Li − cos L0 2

Eb =

 ij VA rij

VR rij − B i

(136)

j>i

where the repulsive pair term is given by e

VR rij = fij rij Dij / Sij − 1 e−



e

2Sij >ij r−Rij

(137)

and the attractive pair term has the form e

VA rij = fij rij Dij Sij / Sij − 1 e−



e

2Sij >ij r−Rij

(138)

The function fij r restricts the pair potential to nearest neighbors and is given by

>  + exp − cos Li

+ rij rik = fc rik exp 63 rij − rik 3

Other Potentials Further expressions for the potential form exist in the case of covalent binding, but here we only want to quote the empirical potential by Brenner [483] which has often been used in the description of numerous processes including the nanometer-scale indentation of diamond surfaces, the deformation of carbon nanotubes, the formation of fullerenes from graphitic ribbons, etc. [484]. The Brenner potential (the so-called hydrocarbon potential) can be formulated as follows [483]: The binding energy for the potential is formulated as a sum over bonds as

(135)

The potential parameters for Si were fitted to the cohesive energy, lattice parameter, and bulk modulus of the diamond structure, and the cohesive energy of bulk polytypes of silicon [480, 481]. In connection with the T3 model an additional constraint was used in order to fit the three elastic constants. The T2 and T3 potentials have been used to study lattice dynamics, microclusters, bulk point defects, the liquid and amorphous states, surface reconstructions, etc. More details are given in [472]. It should be emphasized that the Tersoff formalism is obviously similar to the embedded-atom method (Section 5.3.4) since by a proper choice of functional forms and model parameters the two expressions for the potential energy can be made identical [482].

 1

 = 1 ! r < R  " ij  1

  $ r − Rij

 = 1 1 + cos 2

1

2 fij r

Rij − Rij

   1

2

 Rij < r < Rij    2

 =0 r > Rij

(139)

We do not want to discuss the details of this potential and refer to [483], but the following should mentioned: If Sij = 2, then the pair terms have the form of the usual Morse potential (Section 5.1) and, furthermore, the equilibrium distance e

e

Rij , the well depth Dij , and the parameter >ij are equal to the parameters for the usual Morse potential (independent of the value of Sij ). Critical remarks concerning the Morse potential are given in Section 5.1. Applications The potentials introduced above are important for the description of covalent systems, in particular, in connection with carbon and silicon systems, respectively. Let us quote here some selected studies which are based on the often used potentials by Stillinger and Weber, Tersoff, and Brenner. Stillinger–Weber Potential Articles include: Nearest neighbor considerations in Stillinger–Weber type potentials for diamond [485], Development of an improved Stillinger– Weber potential for tetrahedral carbon using ab initio (Hartree–Fock and MP2) methods [486], Computed energetics for etching of the Si(100) surface by F and Cl atoms [487], Boson peak in amorphous silicon [488], and Strain profiles in pyramidal quantum dots by means of atomistic simulation [489]. Further information is given in [490–499].

Computational Atomic Nanodesign

Tersoff Potential Articles include Modeling chemical and topological disorder in irradiation-amorphized silicon carbide [500], Atomic diffusion at solid/liquid interface of silicon: Transition layer and defect formation [501], Predictive study of thermodynamic properties of GeC [502], Orientation dependence in C60 surface-impact collisions [503], Parameter optimization of Tersoff interatomic potentials using genetic algorithm [504], etc. [505–522]. Brenner Potential Articles include: Monte Carlo simulations of carbon-based structures on an extended Brenner potential [523], Numerical study of shapes of fullerenes with symmetrically arranged defects [524], Energy window effect in chemisorption of C36 on diamond (001)-(2 × 1) surface [525], Reconstruction of diamond (100) and (111) surfaces: Accuracy of the Brenner potential [526], Phonon spectra in model carbon nanotubes [527], and Interaction among C28 clusters and diamond (002) surfaces [528]. More information is given in [529–541].

6. MOLECULAR DYNAMICS In this section we give an overview about recent results in the field of molecular dynamics. Clearly, not all studies could be considered but we selected interesting investigations which, on the other hand, also reflect a broad variety of applications in connection with nanometer-scale molecular dynamics. First a brief overview about various topics is given. Then some specific molecular dynamics models will be discussed which is followed by a more detailed study of metallic nanosystems.

6.1. Design at the Nanolevel 6.1.1. Recent Developments Selected studies about the following topics are compiled: nanomachines, metallic nanosystems, biological systems, nanostructures and nanostructured materials, nanoclusters and nanoparticles, nanotubes and fullerenes, nanowires, and silicon and carbon. Nanomachines Articles include: Molecular mechanics and molecular dynamics analysis of Drexler–Merkle gears and neon pump [542], Docking envelopes for the assembly of molecular bearings [543], dynamics of He/C60 flow inside carbon nanotubes [544], On the importance of quantum mechanics for nanotechnology [545], dynamics and fluid flow inside carbon nanotubes [546], Dynamics of a laserdriven molecular motor [547], and The dynamics of molecular bearings [548]. Metallic Nanosystems Articles include: Simulation of chemical mechanical planarization of copper with molecular dynamics [549], Atomistic simulations of solid-state materials based on crystal-chemical potential concepts: Applications for compounds, metals, alloys, and chemical reactions [550], Metallic nanoislands: Preferential nucleation, intermixing and electronic states [551], Large-scale molecular dynamics study of the nucleation process of martensite in Fe–Ni alloys [552], Molecular dynamics simulation of the grain-growth in nano-grained metallic polycrystals [553], and Melting and freezing of gold nanoclusters [554]. Further studies can be found in [555–594].

77 Biological Systems Articles include: Completion and refinement of 3-D homology models with restricted molecular dynamics [595], Simulation of biological ionic channels by technology computer-aided design [596], Interfacing molecular dynamics with continuum dynamics in computer simulation: Towards an application to biological membranes [597], Structural properties of a highly polyunsaturated lipid bilayer from molecular dynamics simulations [598], Comparison of aqueous molecular dynamics with NMR relaxation and residual dipolar couplings favors internal motion in a mannose oligosaccharide [599], Molecular dynamics simulation of thymine glycol-lesioned DNA reveals specific hydration at the lesion [600], A molecular dynamics study based post facto free energy analysis of the binding of bovine angiogenin with UMP and CMP ligands [601], and Molecular dynamics of DNA quadruplex molecules containing inosine, 6-thioguanine and 6-thiopurine [602]. More information about molecular dynamics in connection with biological systems is given in [603–637]. Nanostructures and Nanostructured Materials Articles include: Molecular dynamics in nanostructured polyimide–silica hybrid materials and their thermal stability [638], Computer simulation of displacement cascades in nanocrystalline Ni [639], Mechanical properties of nanostructured materials [640], Sintering, structure, and mechanical properties of nanophase SiC: A molecular dynamics and neutron scattering study [641], Multiresolution algorithms for massively parallel molecular dynamics simulations of nanostructured materials [642], Computer modeling the deposition of nanoscale thin films [643], and Molecular dynamics simulation of lattice distortion region produced by rounded grain boundary in nanocrystalline materials [644]. Further information about this topic is given in [645–689]. Nanoclusters, Nanoparticles Articles include: Charge dependence of temperature-driven phase transitions of molecular nanoclusters: Molecular dynamics simulation [690], Theoretical study of the atomic structure of Pd nanoclusters deposited on a MgO(100) surface [691], Melting of bimetallic Cu–Ni nanoclusters [692], Structure and properties of silica nanoclusters at high temperatures [693], Platinum nanoclusters on graphite substrates: A molecular dynamics study [694], and Study of self-limiting oxidation of silicon nanoclusters by atomic simulations [695]. More information about molecular dynamics in connection with nanoclusters and nanoparticles is given in [696–731]. Nanotubes and Fullerenes Articles include: Computational studies of gas–carbon nanotube collision dynamics [732], Binding energies and electronic structures of adsorbed titanium chains on carbon nanotubes [733], Microscopic mechanisms for the catalyst assisted growth of single-wall nanotubes [734], Nanotube and nanohorn nucleation from graphitic patches: Tight binding molecular dynamics simulations [735], Carbon nanotubes as masks against ion irradiation: An insight from atomistic simulations [736], Molecular dynamics and quantum mechanics investigation on mechanic-electric behaviors of nanotubes [737], Molecular dynamics study of the fragmentation of silicon-doped fullerenes [738], Fullerene cluster between graphite walls— Computer simulation [739], and Se atom incorporation in

78

Computational Atomic Nanodesign

fullerenes by using nuclear recoil and ab initio molecular dynamics simulation [740]. Further information is given in [741–821].

rotations per picosecond), then slippage occurs but bonds do not break and normal operation will resume when the rotation rate slows.

Nanowires Articles include: Thermal properties of ultrathin copper nanobridges [822], Atomic-scale simulations of copper polyhydral nanorods [823], Size effect on the thermal conductivity of nanowires [824], Structural transformations, amorphization, and fracture in nanowires: A multimillionatom molecular dynamics study [825], and Large deformation and amorphization of Ni nanowires under uniaxial strain: A molecular dynamics study [826]. More information about molecular dynamics in connection with nanowires is given in [827–842].

Bending of Nanotubes The bending of a (8,8) nanotube has been simulated in [923]. The study was performed on the basis of molecular dynamics calculations. For the interatomic interaction Brenner’s potential has been used (Section 5.4.1). The results are shown in Figure 12. The 21 nm long tube contained 2688 atoms. In the simulation a portion of 3.6 Å was fixed from both ends of the tube. One of the ends was forced to move on a half-circular path at a constant rate. The radius of the circular path was a half of the tube length. The total bending time was 250,000 fs during which the simulation temperature was kept at 300 K.

Silicon and Carbon Articles include: Structural transition in nanosized silicon clusters [843], Molecular dynamics studies of plastic deformation during silicon nanoindentation [844], Computational nanotechnology with carbon nanotubes and fullerenes [845], Neck formation processes of nanocrystalline silicon carbide: A tight-binding molecular dynamics study [846], Amorphization and anisotropic fracture dynamics during nanoindentation of silicon nitride: A multimillion atom molecular dynamics study [847], Nonequilibrium molecular dynamics simulation study of the behavior of hydrocarbon-isomers in silicalite [848], Pressurecontrolled tight-binding molecular dynamics simulation of nanoscale nanotubes [849], and Effect of chirality on the stability of carbon nanotubes: Molecular dynamics simulations [850]. Further information is given in [851–921].

6.1.2. Specific Molecular Dynamics Investigations on Nanosystems Nanogears Concerning nanogears interesting models have been studied by Globus and co-workers from the NASA Advanced Supercomputing Division (see, for example, [922]). A typical example is given in Figure 11: The operation of these nanogears was simulated using molecular dynamics calculation and Brenner’s hydrocarbon force field (Section 5.4.1). The study showed that if one gear is turned by an external force, the second gear will turn as well. Furthermore, if the gears are driven too fast (input rate >0 4

Figure 11. On the basis of fullerene nanotechnology it is possible to construct nanomachines, for example, nanogears. A typical gear configuration, designed by A. Globus (NASA Advanced Super-computing Division), is shown in the figure (see, for example, also [822]). The operation of these nanogears was simulated using molecular dynamics calculation on the basis of Brenner’s hydrocarbon force field (Section 5.4.1). Courtesy of NASA.

Atomistic Simulations of Plastic Deformations in Glassy Amorphous Polymers [924] as the demand for polymers with superior properties increases, an understanding of the fundamental connections between the mechanical behavior and underlying chemical structure becomes imperative. In this work, the thermo-mechanical behavior and the molecular-level origins of elastic and plastic deformation in an amorphous glassy polymer were studies using molecular dynamics simulations (Fig. 13a). A velocity Verlet integration scheme with a Berendsen barostat and a Nose–Hoover thermostat were used to study a system of united atom polyethylene. These simulations provide detailed atomic information that has been used to verify experimental results. Stress-induced mobility has been directly observed within these simulations (Fig. 13b), corroborating similar observations in nuclear magnetic resonance experiments of Loo et al. [925]. Dihedral angle transition rates are observed to increase during active deformation and drop off again when deformation ceases. These simulations also show that there is no correlation between local volume and dihedral angle transitions in this system. Models of the Dodecamer DNA Double Stranded Segment Molecular dynamics simulations were performed on models of the dodecamer DNA double stranded segment [d(CGCGAATTCGOG)], in which each of the adenine residues, individually or jointly, was replaced by the water-mimicking analog 2′ -deoxy-7-(hydroxymethyl)deazaadenosine (hmdA). The simulations, when compared to the dodecamer itself, show that incorporation of the analog affects neither the overall DNA structure nor its

Figure 12. Bending of a (8,8) nanotube by molecular dynamics. reprinted with permission from [923] M. Huhtala et al., Comp. Phys. Comm. 147, 91 (2002). © 2002, Elsevier Science.

Computational Atomic Nanodesign

Figure 13. (a) Snapshot of a simulation cell containing a united atom representation of 160 polyethylene chains each with 250 beads. (b) Compressive true stress (red line) and the logarithm of transition rate per bond per nanoseconds, Rtrans (blue line), versus true strain for a system deformed at a constant true strain rate of 5 × 1010 per second along the x-axis at 200 K while the lateral faces were held at atmospheric pressure. Courtesy of F. M. Capaldi, M. C. Boyce, G. C. Rutledge, Massachusetts Institute of Technology (MIT), Department of Chemical Engineering, Cambridge, MA.

79

Figure 14. Molecular dynamics simulations were performed on models of the dodecamer DNA double-stranded segment, in which each of the adenine residues, individually or jointly, was replaced by the watermimicking analog 2′ -deoxy-7-(hydroxymethyl)-deazaadenosine (hmdA). Water molecules within hydrogen bonding distance (proton–acceptor distances < 2 5 Å and angle donor–proton acceptor > 120 to the acceptor/donor groups of the GAATIC bases) were selected and are shown superimposed (for a better understanding the oxygen atoms of the water molecules are shown as vdW spheres with a reduced radius). The DNA atoms are vdW representation. The analog bases are shown in gray. More details are given in the text. This image was made by the Theoretical Biophysics group at the Beckman Institute, Courtesy of University of Illinois at Urbana–Champaign.

we have discussed in Section 5.3.1. In this section we would like to study molecular dynamics results for Al (as a typical example for a metal system); the specific behavior of clusters will be discussed, but also some remarks in connection with

hydrogen-bonding and stacking interactions when a single individual base is replaced by the analog. Furthermore, the water molecules near the basis in the single-substituted oligonucleotides are similarly unaffected (Fig. 14). Double substitution lead to differences in all the aforementioned parameters with respect to the reference sequence [926]. Visualization of the Pathway of Diffusants in Amorphous Polymers The diffusion in nanoconfined polymers has been studied by Boshoff and co-workers [527] in order to study the effects of polymer structure and dynamics on the molecular mechanism of diffusion (see also Fig. 15).

6.2. Metallic Nanosystems As we have outlined, the interaction potentials for nanosystems have to be constructed carefully since surface properties are in general very sensitive to small variations in the potential shape, and in the case of nanosystems a great fraction of atoms belongs to the surface region of the system. This effect is particularly significant in the case of metals. The determination of pair potentials for metals can be done, for example, by means of pseudopotential theory that

Figure 15. A diffusion pathway of a tracer He atom in a glassy, amorphous polypropylene model. Each colored sphere represents a snapshot of the He atom’s position along a 150 ps trajectory (starting at blue and ending at red), as calculated by molecular dynamics at 233 K. Snapshot are spaced 0.1 ps apart. The polymer is represented by carbon–carbon bond for clarity. Courtesy of J. Boshoff, Chemical Engineering, University of Delaware.

80 nanoengineering will be given. The potential for Al used in this study has been determined in accordance with the principles outline in Section 5.3.1.

6.2.1. Interaction Potential for Al In the case of noble gases (for example, krypton [928]) the temperature dependence of system properties can be treated in a good approximation on the basis of temperatureindependent pair potentials. That is in general not possible in the case of metals. Why? As we have already outlined in Section 5.3.1, a metal consists of ions and conduction electrons. The interactions between the ions are strongly influenced by the screening of the conduction electrons, and the screening is sensitive to the density of the conduction electrons, which is, on the other hand, temperature dependent. Therefore, we may conclude that the pair potential in metals is in general dependent on temperature. This property is reflected in the dielectric function  q , which in turn is important for the determination of the pair potential within the framework of pseudopotential theory. How large are such temperature effects for Al? In a first attempt [929] a pseudopotential was used, developed by Dagens et al. [930]. This potential is a nonlocal pseudopotential; the screening is described by the Geldart–Taylor dielectric function. However, this potential was not sufficient to describe certain properties at the surface. Therefore, another model has been chosen for the pseudopotential which is obviously more realistic [931]. Concerning this model, the essential point is the consideration of interactions of van der Waals type in the determination of the direct ion–ion potential as well as in the ion–electron part of the interaction. (Numerical calculations for liquid rubidium near the melting point showed that van der Waals-type interactions are distinctly reflected in the metal potential [932].) The pair potential for Al based on this kind of potential is shown in [933] for two temperatures (300 and 1000 K; the melting temperature is 933 K). The free parameters were fitted with the help of crystal and liquid data for various temperatures and, therefore, the temperature-dependent Al potential describes a wide range of experimental bulk and surface properties: (1) The crystal is stable up to the melting point. The phonon density of states at T = 300 K is described well. (2) The structure and dynamics of the liquid state are also described adequately by the potential: The structure factor (Fourier transform of the pair correlation function) of the liquid state agrees well with experimental data, and the correct value of the diffusion constant is obtained. (3) The melting temperature (933 K) of Al is reproduced by the potential. (4) We found for the outermost layer an inward relaxation, which is confirmed by experiments [934]. (5) The mean-square displacements at the surface agree well with LEED data [934]. (6) The molecular dynamics results indicate that there are relatively strong anharmonic effects already at low temperatures; this is confirmed by experiments [934], too. (7) The onset of premelting (T > 650 K) resulting from the molecular dynamics calculations agrees excellently with the experimental observations.

Computational Atomic Nanodesign

6.2.2. Metastable Cluster-States Molecular dynamics calculations have been performed for cubic nanoclusters with (001) surfaces consisting of N = 500 atoms [935]. For this system the classical Hamilton equations were solved by iteration. The time step in the calculation was 10−15 s. The number of time steps was 4000 (4 ps). The initial values for the positions were chosen to be the bulk structure. In the case of Al and N = 500 particles, the box size of the clusters is approximately 20 Å. The cutoff radius for the interaction potential was chosen as the sixth nearest neighbor distance. The initial velocities have been set according to a temperature of 50 K followed by an equilibration phase of 2000 time steps (2 ps). In the following t = 0 refers to the end of the equilibration phase (i.e., the time at which the system has reached thermal equilibrium). The initial directions of the velocities were distributed randomly, so that the velocity averaged over all particles was zero and remained zero for all time steps. The amplitudes of the velocities determine the temperature T of the system [Eq. (80), Section 4.2.3] and also the 6-function [Eq. (75), Section 4.2.3] which can be used for the study of the thermodynamic state. As already outlined in Section 4.2.3, due to the finite number of particles T and 6 fluctuate around mean values. Figure 16 shows the temperature T and the 6-function as a function of time t. As can be seen, there are three different phases: (1) In the interval 0 ≤ t ≤ 1 ps the cluster remains in thermal equilibrium (6 fluctuates around 5/3) at a mean temperature of 50 K (with the usual, stable fluctuations). (2) Within the interval 1 ≤ t ≤ 2 ps a transition takes place without any external influence, and within this time interval the temperature rises up to 160 K. (3) Finally, for t > 2 ps

Figure 16. Temporal development of the temperature and the 6-function for a cubic Al500 cluster with (001) surfaces. Initially, the cluster remains in thermal equilibrium at about 50 K for 1 ps (6 = 5/3 indicates thermal equilibrium, Section 4.2.3). Then, during a transition phase the cluster temperature increases to 160 K. During the transition phase of 1 ps the cluster leaves equilibrium. After this process, the mean temperature remains at 160 K for all times and the system also remains in thermal equilibrium. It should be emphasized that the whole process takes place without external influence; that is, there is no outer effect causing the transition.

Computational Atomic Nanodesign

81

the cluster temperature remains at 160 K and in thermal equilibrium (6 fluctuates around 5/3). How can we interpret this transition? What is the driving force behind this phenomenon and, furthermore, under what conditions does it occur? The cluster is obviously in a metastable state within the first phase (50 K), and in the third phase (160 K) it is in a stable state. Further, due to the energy conservation of the system the potential energy of the cluster in the stable state has to be smaller compared to that in the metastable state. Therefore, the structure of the metastable state should be different from that of the stable state, since the potential energy depends on the relative distances of the atoms. Let us first study the structure and after that the Dynamics of the cluster in the metastable state as well as in the stable state. Structure The structure of the different cluster states has been studied by means of the structure factor (see, for example, [393]) S q =

N 1  expiq ri − rj N i j=1

(140)

where · · · denotes again statistical averaging, q is the wave vector, and ri is the position vector of the ith particle. By averaging of q over all directions, S(q) only depends on the magnitude of q: S(q with q = q . Figure 17 shows the results of S q for the metastable and the stable state together with the results for the bulk state (500 particles with periodic boundary conditions). As can be seen, the typical double peak of fcc-structured materials at about 2.7 and 3.1 Å−1 appears also for the other states (indicated in Fig. 17 by gray bars). However, the metastable cluster state additionally exhibits a well pronounced peak at 3.4 Å−1 which is missing in the stable state as well as in the bulk (indicated by the yellow bar). This peak can be interpreted as a superstructure that has been formed within the metastable cluster due to a periodical particle shift. Dynamics The cluster dynamics has been studied by means of the generalized phonon density of states f 7 , and this quantity can be calculated from the velocity autocorrelation function - t by means of a Fourier transform (Section 4.2.3). As already outlined in Section 4.2.3, the Fourier transform of - t yields a frequency spectrum which is, in the case of the harmonic solid, the frequency spectrum of the normal modes. Thus, for an anharmonic system, for example, clusters, we can define a generalized phonon density of states f 7 by means of the Fourier transform of - t . - t has been calculated by molecular dynamics, and its Fourier transform is shown in Figure 18. In Figure 18 the generalized phonon density of states is given as a function of  with 7 = 2$. For the metastable state f  at 50 K is quite different from that of the stable state at 160 K. In particular, the pronounced peak at 2 THz (metastable state) does not appear in the frequency spectrum of the stable state (and also not in the bulk). Thus, we may conclude that the appearance of the superstructure in the metastable state (Fig. 17) is accompanied with the appearance of the low frequency mode at 2 THz. This mode is completely missing

Figure 17. Structure factor S q of a cubic Al500 cluster with (001) surfaces which is in the metastable and stable state, compared with S q

for the bulk (also here 500 particles have been used but with periodic boundary conditions). The light blue bars indicate peaks which are common for all phases; they are typical for Al. The yellow bar emphasizes the well pronounced peak at 3.4 Å−1 that appears only in the metastable phase and could be a hint for a superstructure.

Figure 18. Generalized phonon density of states f  of a cubic Al500 cluster in the metastable and stable state. The well pronounced peak at 2 THz in the frequency spectrum of the metastable state is recognizable to some extent only in the stable state. This peak results from surface oscillations at the metastable cluster.

82 in the spectrum of the solid in the bulk as well as in the liquid bulk state. The low frequency mode is probably the cause for the transition from the metastable to the stable cluster state. Let us briefly discuss this. The outer dimensions of the metastable cluster are smaller compared to the initial bulk-structured cluster. Therefore, during the equilibration phase the whole cluster passes through a “shrinking process,” and due to the repulsion, this shrinking process should be followed by a “recoiling effect”; that is, there is an oscillation reflected in a temporal change of shrinking and recoiling leading to a “breathing” of the whole cluster. In connection with Figure 16 we discussed that the potential energy Vms of the metastable state must be larger than the potential energy Vs of the stable state. Since the transition takes place after a certain time and not at once, we can conclude that there is a certain potential barrier EVms which the cluster has to overcome. The “breathing” of the cluster can lead to the effect that the system has, at a certain time, the potential energy V ≥ Vms + EVms , and then the transition from the metastable to the stable state can take place; it is only a question of time until a fitting configuration has established by chance which yields a potential energy high enough (V ≥ Vms + EVms to overcome the potential-energy barrier. Analogies to Atomic Excitations There is a certain analogy between metastable states of clusters and of atoms in an excited electron state: (1) The potential energy in the excited (metastable) state is larger than in the ground (stable) state. (2) The systems (atoms, clusters) leave their excited (metastable) state without external influence after a certain time interval: the atoms by emiting photons, the clusters by increasing their temperature (see also Fig. 16). The atoms are in an excited state for 10−8 s while the clusters are in a metastable state for approximately 10−12 s. The time interval of 10−12 s is much smaller than 10−8 s (excitation time of atoms). Nevertheless, the lifetime of 10−12 s of the metastable state is relatively large; it is distinctly larger than the time interval (0 4 × 10−12 s) that a sound wave needs for the propagation across the nanosystem. The lifetime of such metastable states is a function of the particle number and is also a function of the outer shape of the nanosystem [3].

6.2.3. Bifurcation Phenomena In connection with metastable states of clusters the following effect is of particular interest: When a cluster transforms from the metastable to the stable state, there are several possibilities for that. In other words, the local minimum in the potential energy of the metastable state is surrounded by numerous deeper minima which correspond to numerous stable states. Which of these stable states the cluster will finally take cannot be predicted in principle. Due to the unavoidable fluctuations in the potential energy of such nanosystems it is completely a matter of chance (Section 6.2.2). In other words, in connection with cluster state transitions there is a bifurcation in the sense of chaos theory (Fig. 19): At the bifurcation point nature plays dice to decide on which

Computational Atomic Nanodesign

Figure 19. Illustration of the bifurcation phenomenon: In principle it cannot be predicted in what stable state the metastable systems transits. At the bifurcation point nature plays dice in order to decide on which of the various branches the cluster will finally rest.

of the various branches (stable states) the cluster will finally rest. Stable cluster configurations may differ in both the inner structure as well as the outer shape. In general, the inner structures show multicrystalline compositions (i.e., there are grain boundaries, dislocations, and other lattice defects). The outer shapes of the clusters strongly depend on the arrangement of these lattice defects. This is an interesting phenomenon and there is no counterpart in micro- and macrotechnique. In macrotechnique, a system (for example, a machine) will always be destroyed after a certain time, if the mechanic does not keep the machine in good condition; this degradation of the system is due to its interaction with the systems surrounding. There are no other transitions in micro- and macrotechnique. As we have seen, the situation is different in connection with nanosystems: The transition at the bifurcation point is constructive; the system will not simply be destroyed but there emerge new forms, namely one of the possible stable states (see Fig. 19). As said, in micro- and macrotechnique we do not observe such phenomena, and this is because there are no such metastable states. If there were, nevertheless, metastable states also established in the macroscopic realm, in principle a screwdriver could transform spontaneously into a hammer or a nippers; Nature plays dice; a screwdriver becomes either a hammer or a nippers, and such transitions would take place without any external influence. We know that such phenomena are not possible in micro- and macrotechniques but should be real phenomena in nanotechnology, in particular, within the frame of materials research at the nanolevel. The bifurcation phenomenon is the behavior of single clusters. This specific behavior comes into play by the mutual influence of temperature and the interaction between the particles forming the cluster. There is a certain kind of independent creativity which does not come from outside; that is, it is not forced by the observer or by other factors of the

Computational Atomic Nanodesign

environment but is an inherent characteristic of the system itself. Such behavior is also known in connection with biological systems, which are, however, much more complex compared to the Al clusters studied here. Since this specific behavior (creativity) is essentially influenced by the temperature and the interaction potential between the particles, nanosystems at zero temperature have to be considered as “dead” systems without any creativity.

6.2.4. Final Remarks Although the Al clusters represent inorganic systems there are certain analogies to biological systems: (1) A nanosystem can be creative; that is, it can meet individual decisions which cannot be influenced from outside; the bifurcation phenomena (Fig. 18) can be interpreted in this way. (2) A nanosystem is able to transform spontaneously; it can transform from a metastable state to a stable state without external influence. Spontaneous transformations are also typical phenomena within biology, in particular in connection with embryology. (3) New forms, new shapes, which were not existent before, emerge without external influence. Also within biology (e.g., in embryology) new forms appear, which were not existent before. (4) The shapes of such nanosystems transform to more complex forms. This behavior too is typical for biological systems and is a feature of biological evolution. The nanosystems transform, for example, from a cube to a cuboid, and a cuboid is more complex than a cube because one needs more description elements to characterize it.

6.3. Nanoengineering 6.3.1. Nanomodels So far the general procedures and different elements for designing molecular dynamics models have been discussed. As long as there are just single and simple shaped objects of interest, like cubical nanoclusters, the setup of an appropriate molecular dynamics model is relatively easy. Since the initial positions are given by the perfect lattice structure such objects can be built by using any programming language to implement the according algorithms. Depending on the variety of materials, characterized by the lattice constant and structure (fcc, body-centered cubic, hexagonal, etc.), and depending on the possible orientations [(001), (011), (111), and so on] the software development is more or less extensive. For the generation of (semi-infinite) surface models the additional application of periodic boundary conditions (within two dimensions) is necessary. But obviously such simple objects are only of minor interest within the field of nanoengineering, where either structures with complex shapes or interactions (cluster–cluster or surface–cluster) under various conditions play an important role. Especially for studies of nanomachines (but for others too) the molecular dynamics (MD) models have to contain moving parts in addition to objects of miscellaneous materials, and therefore, the software should be able to combine single structures, each with different initial values.

83 Quite clearly, the development of a computer program that can generate MD models for all the above-mentioned topics is rather extensive. While the setup of simple structures (cuboids, spheres, surfaces) and the setup of the initial values according to rotational or transitional movements and for the construction (positioning, centering, adjusting) of individual objects may be fixed by usual programming, the nanodesigning part of complex shapes has been accomplished by a common commercial software. The idea arose from the fact that for metallic nanostructures—unlike covalent bonded materials—there are basically no restrictions concerning their initial shapes (i.e., in this case a molecular dynamics model can be built first by defining the outer shape and then by filling it up with atoms according to the crystal lattice). Since the design of three-dimensional shapes is a very common topic within mechanical engineering—the so-called computer aided design (CAD)—one can use a suitable CAD software with a programming interface to perform such a task. Here the choice fell on the commercial software package Genius Desktop (in combination with Mechanical Desktop ) that includes a LISP derivative as programmable interface. At first, two additional system functions have been developed, one to compute the outer limits of a threedimensional CAD object, the other to determine whether a given spatial point lays within a selected object or not. Then, with the help of these functions the implementation of various filling, scaling, and storage algorithms for different lattice structures has been performed. But in contrast to common mechanical engineering the nanodesign does not end here. The nanomodel with its initial atomic perfect lattice positions corresponds to a temperature of 0 K. So the next step is to accelerate the atoms until an equilibrium is reached according to a prescribed specific model temperature. Such an equilibration is performed during a sufficient number of MD calculation steps by rescaling the particle velocities: vi t ←

#

Td v t

T i

i = 1 N

(141)

where Td is the desired model temperature, and T and vi are the current temperature value and velocity vectors. During such an isothermal equilibration phase, the rescaling needs to be applied after about each tenth time step only. The rescaling of velocities is not only useful for the equilibration of nanomodels but can be applied during molecular dynamics calculations too, where varying Td with time leads to artificial tempering or cooling processes.

6.3.2. Differences Between Nanoengineering and Common Mechanical Engineering Most often the initial molecular dynamics model, according to the perfect lattice structure, is not a stable configuration. Figure 20 illustrates the typical design steps in the case of metallic nanostructures. First, a CAD model is established which defines the rough layout, known from common mechanical engineering. Then the molecular dynamics model is created by filling the CAD design with single atoms.

84

Computational Atomic Nanodesign

calculation, and redesign, and there is almost nothing in common with common mechanical engineering.

6.3.3. Nanomachines The synthesis of real nanomachines has not yet been performed, but it is a most interesting topic to perform molecular dynamics studies and, therefore, to determine how and under what conditions nanomachines could be working— regardless of the question of how to assemble them in reality. An example of such a study is illustrated in Figure 21. This is a model of a simple nanoturbine that consists of a paddle wheel with an axle and two bearings resting on a substrate. The propulsion of the turbine could be established by a particle or laser beam. When comparing Figure 21a with Figure 21b the differences between common mechanical engineering and nanodesign again become obvious. While mechanical engineering

Figure 20. The three steps of nanodesign: First, a CAD model has to be established. Then, the molecular dynamics model is created by filling the CAD object with 1958 aluminum atoms. Finally, temperature is applied on the model during an equilibrium phase (in this case 300 K).

Finally, the realistic nanostructure results from the equilibration process which applies a prescribed temperature to the molecular dynamics model. As can be seen quite clearly, the final shape differs significantly from the CAD model. How can this be explained? The atomic positions of the molecular dynamics model are set according to the perfect lattice structure. While such a structure is a stable configuration in macroscopic realms this is no longer valid for nanometer-sized objects. Therefore, the atoms of the molecular dynamics model rearrange their positions during the equilibration phase until a stable configuration is reached. Most often this leads to quite different results compared to the initial design. In conclusion, one has to recognize that, in comparison to common mechanical engineering nanodesign strongly depends on the outer shape of the considered structure. Another difference is significant too: While in mechanical engineering sharp edges or plane surfaces or any shape and also continuous calibrations are common, nanodesign is limited to certain positions of single atoms (i.e., the shape of structures is restricted to certain measures dependent on the material). This becomes obvious, if one thinks about such trivial mechanical parts like fittings and bearings. And in addition, the initial shape of a nanomodel which may have more or less pronounced edges can change considerably during the equilibration phase (as well as later on during the molecular dynamics calculation) and, therefore, the layout of the model would have to be redesigned. That is, building nanomodels or performing nanoengineering is an iterative process of nanodesign, molecular dynamics

Figure 21. Nanoengineering of a simple turbine. The turbine consists of a paddle wheel with an axle and two bearings. (a) Mechanical engineering model build with common CAD software. (b) Perspective view and cross-section of the MD model [the corresponding side view is shown in (c)]. The paddle wheel rotates with 5 × 1010 revolutions per second. (d) The paddle wheel ruptures, if the number of revolutions is increased to 1011 per second model [the side view of (d) is shown in (e)]. The bearings (blue) consist of 4592 silicon atoms, each with ideal shaped surfaces. Due to the strong covalent bonds of silicon the bearings have been treated as rigid objects within the MD calculations. The axle with the attached paddle wheel (yellow) has been assembled with 5226 aluminium atoms and the substrate (green) consists of krypton.

Computational Atomic Nanodesign

usually works with sharp edges and smooth surfaces, nanodesign is restricted to the use of single atoms which leads to the typical grainy appearance of nanostructures. Moreover, tolerances may be kept very small with mechanical engineering, but nanodesign allows measures only in certain steps where the steps depend on the according material. Therefore, the fit of axle and bearing is rather loose as can be seen in the cross-sections of Figure 21. However, the MD studies have shown that such nanoturbines could be working stable for revolutions of up to about 5 × 1010 per second.

7. MONTE CARLO METHOD In contrast to molecular dynamics calculations, where we have completely deterministic algebraic equations, the Monte Carlo method is a numerical approach in which specific stochastic elements are used. The Monte Carlo method was introduced by Metropolis et al. [936] in 1953. The method and results have been extensively reviewed in the literature (see, for example, [937]). Since we have concentrated in this chapter mainly on molecular dynamics, we only want to discuss here some basic characteristics. Whereas the molecular dynamics method enables the investigation of static and dynamic properties, the Monte Carlo method only deals with static properties; these are configurational averages and thermodynamics quantities. The results produced by the two methods should be in agreement to order N −1 and, therefore, one can expect agreement of the two methods within statistical error. It should be emphasized that the Monte Carlo method may be easily extended to quantum mechanical systems. The widespread applications of the Monte Carlo method made it to an important tool in science and technology. In connection with condensed matter physics typical applications are (see, for example, [937]): fluids and plasmas, thermal properties and scattering functions for solids, liquids, and dense gases; properties of metallic alloys (in particularly in connection with short- and long-range-order properties); relaxation phenomena and diffusion processes; kinetics of crystal growth; phase transitions; magnetic properties of surfaces; kinetics of adsorption on surfaces; structural and thermal properties of glasses and amorphous magnets and of other disordered systems; ground state properties of certain quantum systems; etc. The Monte Carlo method has been used in a diverse number of ways. For example, in connection with molecular calculations five types have often been used: classical Monte Carlo, quantum Monte Carlo, path-integral quantum Monte Carlo, volumetric Monte Carlo, and simulation Monte Carlo. But there are still many variations of these five types in literature. Here we only want to give the main principles of the classical Monte Carlo method and those of the quantum Monte Carlo method.

7.1. Classical Many-Particle Problems The Monte Carlo method enables the determination of configurational averages. Let us consider a quantity 6 which depends on the configuration T = T r1 rN of the N

85 particles of the system: 6 = 6 T . The configurational average of 6 is given in a canonical ensemble by 6 = where



dT . T 6 T



exp − Vk TT

B . T = 

dT exp − Vk TT

(142)

(143)

B

The Monte Carlo method enables the determination of integrals of type (142). Below we discuss the main points of this procedure. Let us split the configurational space into k cells with sufficiently small elements ET , and let Tm , m = 1 k, be the corresponding configurational states. Then, the average (142) may be approximated by 6 MC =

k



6 Tm exp − Vk TTm

B V Tm

k m=1 exp − k T

m=1

(144)

B

We can estimate (144) by selecting at random a certain number L 10 nm for Co). On the other hand, hard ferromagnetic materials, which often exist as alloys, exhibit high anisotropies. Soft ferromagnetic materials embedded in a noble metallic matrix show a technologically important effect called giant magnetoresistance (GMR). The ability to prepare air-stable ferromagnetic/noble metal bimetallic particles is therefore a challenge. Well-known ferromagnetic alloys consist of ferromagnetic 3d and noble metal elements of the platinum group. They are known to have large magnetic moments and large magnetic anisotropies [74]. Accordingly, the synthesis of monodisperse ferromagnetic/noble metal bimetallic particle alloys (especially Fe/Pt) or coreshell particles and their assembly into ordered arrays have been investigated intensely in recent years. Rivas et al. [75] reported the synthesis of cobalt nanoparticles coated with silver, based on the successive reduction of Fe2+ or Co2+ and Ag+ within the aqueous microdroplets of inverse microemulsions. A similar approach was later used by Seip and O’Connor [76, 77] for the synthesis of Fe@Au using cationic surfactants, with magnetic measurements showing superparamagnetic behavior. Klabunde and co-workers [78] reported the formation of core-shell structures upon heat treatment of metastable alloy nanoparticles of immiscible metals prepared through the so-called solvated metal-atom dispersion (SMAD) method. For the case of Fe-In and Fe-Nd, but with Fe-Au, Fe-Bi, Fe-Ca, Co-Mg, and Ni-Mg, either stable alloys or segregated particles were formed. Magnetic studies of the Fe@In and Fe@Nd samples showed low coercivities and Fe crystallite size-dependent saturation magnetization, though apparently there was always an increase in stability against core oxidation. More recently, the synthesis of well-defined, core-shell Co-Pt nanoparticles was reported by Park and Cheon [79]. The formation of these alloys is driven by redox transmetalation reactions between the reagents without the need for any additional reductants. While the reaction between Co2 (CO)8 and Pt hexafluoroacetylacetonate results in the formation of alloy nanoparticles such as CoPt3 , the reaction of Co nanoparticles with Pt hexafluoroacetylacetonate in solution results in Co@Pt morphologies, which is schematically shown in Figure 4. Core-shell nanoparticles with an inverted morphology were prepared by Sobal et al. [80], with Ag cores coated with uniform Co shells. These authors report an unexpected stabilization of the Co shell against oxidation, which can be due to electron transfer from the core. In this case, the higher electron density of the core allows a sufficient contrast in TEM to distinguish the core-shell morphology (Fig. 5).

2.5. Silica-Coated Nanoparticles We have decided to discuss silica-coated nanoparticles separately because of their special interest, which is also the reason for the extensive work that has been performed on

203

Figure 4. Synthetic routes of core-shell and alloy nanoparticle formation via transmetallation reactions. Reprinted with permission from [79], J.-I. Park and J. Cheon, J. Am. Chem. Soc. 123, 5743 (2001). © 2001, American Chemical Society.

these systems. One of the major reasons for silica coating is the anomalously high stability of silica colloids, especially in aqueous media, but other reasons are the easy control on the deposition process (and therefore on the shell thickness) and its processibility, chemical inertness, controllable porosity, and optical transparency. All of these properties make silica an ideal and costless material for tailoring surface properties, while maintaining the physical properties of the underlying cores. This technique was pioneered in the late 1950s by Iler [81], who described a method for coating with silica “particles with at least one dimension which is less than about 5 microns,” and the nature of which is typically metal oxides and silicates. Silicate solutions with suitable concentration, pH, and temperature were termed active silica, which easily sticks to the surface of oxide colloids in suspension. The slow reaction ensured deposition of thin silica shells, thus retaining the surface shape of the core particles. An alternative to Iler’s silica coating procedure was developed by Ohmori and Matijevic [82], following the already classical Stöber method [83] (base catalyzed hydrolysis and condensation of tetraethoxysilane (TES)) to coat hematite (Fe2 O3  spindles in ethanol. This procedure was later modified by ThiesWeesie [84]. Later work by Philipse et al. used a combination

Figure 5. TEM image at low magnification of a monolayer of Ag@Co nanoparticles deposited on a carbon-coated copper grid by drying of a drop of solution. The core-shell structure is visible at higher magnification (see inset). Reprinted with permission from [80], N. S. Sobal et al., NanoLetters 2, 621 (2002). © 2002, American Chemical Society.

204 of the previously described procedures for coating magnetite (Fe3 O4  superparamagnetic nanoparticles [85] and boehmite (AlOOH) rods [86]. Following Philipse’s work, the silica coating of magnetite was later modified independently by Liu et al. [87] and by Correa-Duarte et al. [88], both achieving more homogeneous and well-defined shells. The control of magnetic interactions is an important consequence of the coating of magnetic particles, which was demonstrated by Donselaar et al. for particles in solution [89] and by Aliev et al. in close-packed thin films [72]. A different approach was followed by Schmidt and co-workers, who used various organosilanes as building blocks and spacers to introduce functionalities within the nanoparticle structure, which can be subsequently employed for preparing more complex materials [90, 91]. Silica coating of metal colloids was also studied in detail, specially during the last decade. This type of core-shell particle presents as an additional difficulty the chemical mismatch between the core and the shell materials. Thus, a procedure must be devised to link the core and shell materials with each other. Several routes have been followed with better or worse results. Ohmori and Matijevic [92] prepared Fe@SiO2 particles by an indirect path comprising the initial coating of hematite spindles [82], followed by reduction of the core within a furnace in the presence of hydrogen. The high temperature promoted sealing of the pores of the silica shell, thus avoiding further oxidation of the metallic core. Liz-Marzán and Philipse [93] first approached the synthesis of Au@SiO2 particles through the formation of nanosized gold particles on the surface of small silica spheres, followed by extensive growth in ethanol, so that Au was embedded as a core. However, a large proportion of pure silica spheres was also produced, which were difficult to separate. A method that has provided substantially better results was later designed by Liz-Marzán et al. [94, 95], with the use of bifunctional molecules to anchor silanol groups on the nanoparticle surface, followed by a slow deposition of a thin silica layer in water, and subsequent growth in ethanol, resulting in monodisperse colloids with one Au core at their center and an outer silica shell of controlled thickness. An example of the concentric geometry achieved is shown in Figure 6. This system proved ideal for a systematic study of optical properties [95, 96]. The same method was later extended to Ag@SiO2 by Ung et al. [35], though dissolution of the Ag cores was observed when concentrated ammonia was added to increase the thickness of the silica shell. The same team [35, 97–98] demonstrated that dissolution by ammonia is due to the possibility of chemical reactions taking place on the cores because of the porosity of the shells.

Core-Shell Nanoparticles

Another advantage of this method is that the first deposition in water is so slow that the shape of the original core is preserved in the initial stages, which was shown first by Chang et al. [99] and then by Obare et al. [100] by coating gold nanorods with the same procedure. The method was also modified by Hall et al. [101], who used silane coupling agents during the growth of the shell to obtain particles with amino groups on the outer surface. Hardikar and Matijevic [102] demonstrated that for larger silver particles (60 ± 5 nm) stabilized with a complicated compound (Daxad 19), no coupling agent is necessary for silica coating through TES hydrolysis in 2-propanol, and this was also recently shown by Xia and co-workers for PVP-stabilized Ag nanorods [103]. Mine et al. have also recently found [104] that standard, citrate-stabilized Au nanoparticles can be homogeneously coated with relatively thick shells without the use of a silane coupling agent, through careful tailoring of the concentrations of ammonia and TES, as well as choosing the proper addition sequence. This leads to a cleaner system (no chemical alteration of the metal surface), but almost invariably implies the formation of some core-free silica particles in addition, which is undesirable for some applications. Ostafin and co-workers have recently devised a procedure to create mesoporous silica shells on Au nanoparticles coated with a thin silica layer by using surfactants during shell formation [105]. The same method was also applied [106] to coat CdS nanoparticles, observing that the silica shell not only enhanced the colloidal stability of the quantum dots, but also the photostability, so that no photodegradation took place, even under light and in the presence of oxygen. A similar procedure was used by the Alivisatos group [107, 108] with the aim of using highly luminescent quantum dots coated with silica for biological labeling. A new variation of the same method was used by Rogach et al. [109] for the coating of CdTe and CdSe@CdS nanocrystals, which resulted in the inclusion of multiple cores within every silica sphere, as well as a dramatic drop in luminescence intensity upon coating. Other methods were later designed for the preparation of Ag@SiO2 . The group of Adair [110] reported the coating through consecutive reactions within microemulsion droplets. The problem with this technique is that removal of surfactants is expensive, tedious, and time consuming. Pastoriza-Santos and Liz-Marzán prepared Ag@SiO2 particles through reduction of Ag+ by N N -dimethylformamide (DMF) [111] in the presence of an amino-silane. High temperatures were needed to achieve fast reduction of the silver salt, prior to condensation of the organo-silica shell, but rather monodisperse silver cores were obtained, and the particles could be extracted into water or ethanol.

2.6. Semiconductor Shells on Metal Cores

Figure 6. TEM micrographs of Au@SiO2 nanoparticles with constant core size and increasing shell thickness, grown by TES hydrolysis.

Research has also been carried out to deposit other oxide shells on nanoparticles, particularly on metal nanoparticles. In this way, it is intended to make use of semiconducting properties of some oxides, such as TiO2 or SnO2 , in combination with the conducting properties of the core. This can be exploited for biological and electronic applications, as detailed below.

Core-Shell Nanoparticles

One example is the synthesis of Ag@TiO2 by PastorizaSantos et al. [112] through the reduction of AgNO3 in DMF/ethanol mixtures, in the presence of titanium tetrabutoxide, which condensed on the surface of the silver cores. Layer-by-layer assembly of the resulting core-shell nanoparticles, followed by dissolution of the silver cores with ammonia, subsequently led to the development of ion-selective and biocompatible titania nanoshell films [113], which were found to be useful for monitoring the diffusion of dopamine, an important compound for neurochemical processes. Coating of gold nanoparticles with titania was also achieved by Mayya et al. [114], who used a different TiO2 precursor (titanium(IV) bis(ammonium lactato) dihydroxide), which is negatively charged and readily complexes with a positively charged polyelectrolyte, poly(dimethyldiallyl ammonium chloride), previously assembled on the nanoparticle surface, and can then be hydrolyzed, providing a good control on the morphology of the shell. The same authors applied this procedure to coat Ni nanorods with titania [115], and upon dissolution of the Ni cores they were able to obtain uniform TiO2 nanotubes. Typical electron micrographs of the core-shell nanorods and titania nanotubes are shown in Figure 7. Another interesting application of metal nanoparticles surrounded by semiconductor shells is the fabrication of composite nanoparticles with a large electronic capacitance. The idea is that there is a large difference between the Fermi level of the core and the conduction band energy of the shell, so that electrons diffusing through the shell can be trapped in the core for a long time. Mulvaney et al. [116] explored this possibility with Au nanoparticles encapsulated in a polycrystalline SnO2 shell. These authors demonstrated that charge can be injected by cathodic polarization through -radiolysis, which resulted in a blue-shift of the plasmon

Figure 7. TEM images of (A) a nickel nanorod coated with titania, (B) titania-based nanotubes obtained by dissolution of coated nickel nanorods, (C) a higher magnification image of a nanotube. (D) A high-resolution image showing the presence of titania nanoparticles. Reprinted with permission from [115], K. S. Mayya et al., Chem. Mater. 13, 3833 (2001). © 2000, American Chemical Society.

205 band (as predicted by theory), which remained blue-shifted in the absence of oxygen, but red-shifted back when open to air because of oxygen reduction by the electrons stored in the core. We finally include in this subsection metals surrounded by semiconductor shells which are not composed of oxides, but rather of chalcogenides. Kamat and Shanghavi [117] prepared Au@CdS nanoparticles by surface modification of Au colloids with mercaptonicotinic acid followed by the addition of Cd2+ and exposure to H2 S. The result was the formation of very small CdS particles attached to the surface of the Au cores, but also free in solution. To prepare particles with novel optical properties due to both local field effects of the metal and further enhancing nonlinear optics of quantum dots, Nayak et al. [118] generated Au@CdSe particles in TOP/TOPO organic solvents, though a large proportion of particles were a mixture of individual Au and CdSe particles, as detected by TEM. Hirano et al. [119] encapsulated nanosized gold with boron nitride at extremely high temperatures and proposed that these particles would have potential applications for single-electron transistors.

3. INORGANIC–ORGANIC CORE-SHELL NANOPARTICLES Principally inorganic-organic nanoparticles can be divided into systems with an organic core and an inorganic shell or vice versa. Although this differentiation seems to be necessary for the specific properties of the prepared materials, similar methodologies are often used for their synthesis. A prominent technique for the preparation of such cores is the use of microemulsions [2, 120]. In this approach, in which the dispersed stabilized phase can be considered as a nanoreactor, the size, shape, and curvature of the surfactantstabilized phase determine the shape of the nano-objects obtained. With this technique the core and the shell are formed with the use of the same microemulsion by simply subsequently building up the two moieties in the micelles. In a different approach the core is formed first, isolated, and modified with the shell in a separate step by covalent anchoring or adsorption of organic groups on the surface. For the synthesis of defined core-shell systems it is crucial to obtain in a first step well-defined cores. Water-inoil microemulsions, sometimes also called reversed micelles, are probably the most widely investigated emulsion systems for the preparation of spherical inorganic nanoparticles. The metal precursor is usually dissolved inside the water droplets and reacts in this confined arrangement to the final particle. Two different methods were used for the formation of the nanoparticles: either all chemicals for the nanoparticle formation are already included in the water phase and the metal precursor is added, or two water-in-oil microemulsions, both containing precursors for the nanoparticle formation, are mixed. In the latter case, the exchange of substances occurs in the united microemulsion by diffusion and droplet collision. With the microemulsion method a variety of nanoparticles were synthesized, such as SiO2 [121–124], TiO2 [125–128], ZrO2 [129], CdS [130], magnetite [131], or mixed metal oxide systems [132]. Surfactants are crucial for the stabilization of the microemulsions; they produce compatibility between the two

Core-Shell Nanoparticles

206 phases and act as barriers against the agglomeration of the final particles. These molecules cover the surface of the particles formed and often interact with it, which can directly affect the local structure of the surface [133]. If the particles have diameters in the lower nanometer range, this interaction can result in a significant change of the particle properties. The capping of the surface with the surfactant is frequently used for the stabilization of the particles formed. With this methodology, for example, surface-capped Pd [134], Fe2 O3 [135, 136], and CeO [137] particles were formed. Emulsion polymerization has the potential to produce a stable and compact polymer layer, completely enclosing the particle and improving its chemical resistance. This is particularly important for pure metal particles that have a tendency to undergo oxidation reactions. Changing the concentration of monomers can readily control the thickness of the layer, which usually protects the core very efficiently. With this method, for example, Ag nanoparticles were encapsulated in styrene and copolymers of styrene and methacrylate. Encapsulated particles retain their optical properties and display a remarkable chemical stability [138]. Dense silica nanoparticles can be prepared by the Stöber process, as already mentioned [83]. The size of the nanoparticles formed can be controlled over a wide range with a narrow size distribution. Therefore, this method is often the source for dense silica nanoparticles, which can be subsequently surface modified after isolation. Additionally, the formed alcosol can also be used as a reaction medium for the direct surface modification of the particles, for example, with trialkoxy-substituted silane coupling agents (Fig. 8) [139]. Depending on the coupling agent, a change of the particles’ properties, such as their solubility, coagulation behavior, or their reactivity, can be observed. Furthermore, the use of organically modified trialkoxysilanes in the synthesis of the silica core provides a method for decreasing the crosslinking density and/or to incorporating additional functionalities, such as dyes, into the core [90, 140]. For example, the use of MeSi(OR)3 and Me2 Si(OR)2 instead of Si(OR)4 can subsequently decrease the cross-linking density of the particle core; therefore these particles can mimic a behavior, for example, with respect to solvent uptake that is between those of pure silica and cross-linked polymers [90]. Polymer chains were grafted to organosilicon nanoparticles that carried a Si-H surface modification [141]. PS with a terminal unsaturated bond was reacted with these particles by the application of a hydrosilation reaction. With this method particles with a radius of 10 nm were reacted with PS chains with gram molecular weights up to 20,000 g/mol. The resulting core-shell particles reveal a high compatibility with linear PS polymers if the molecular weight of the free polymer is equal to or smaller than that of the grafted polymers. In these cases, an optical transparent film

is obtained, while the pure organosilicon particles are totally incompatible.

3.1. Metal Nanoparticles with Capping Ligands Probably the oldest nanoparticle system known is that of gold colloids, which were already used hundreds of years ago for the coloring of window glass. Because of their relatively simple preparation and exceptional stability, gold nanoparticles have also become an important research issue in the present time. In addition, gold nanoparticles are particularly easy to modify because they are often stabilized with weakly binding layers of charged ligands (e.g., citrate) that can be replaced with molecules with chemical functionalities that bind more strongly (e.g., thiols, amines, and disulfides) to their surfaces than these ligands. A twophase synthesis approach results in monodisperse, stable, and soluble gold nanoparticles with a core size of a few nanometers [142]. The synthesis is based on the transfer of AuCl− 4 from an aqueous phase into an organic phase under phase transfer conditions followed by reduction in the presence of thiols. This methodology produces thiol-capped gold nanoparticles. If the thiol employed has additional reactive functionalities attached, the surface-modified nanoparticles can undergo chemical reactions [142]. Another advantage of thiol-capped gold nanoparticles is that the ligands can undergo ligand-exchange reactions with other thiols [143]. The thiol-stabilized Au nanoparticles showed spontaneous self-assembly phenomena depending on the surface modification [144]. Mixtures of different thiol-capped metal cores (i.e., Ag and Au) were used for the formation of ordered two-dimensional arrays, so-called nanoalloys [145]. More information on aspects of the self-assembly of nanoparticles can be found in some recent reviews [146, 147]. Similar to the gold nanoparticles, other metals and semiconductors can serve as cores and bind organic molecules with anchor groups, allowing a strong interaction with the metal surface. Among others, platinum nanoparticles were prepared with the use of alkyl isocyanides as capping ligands, or the self-assembly of films of CdS nanoparticles was enhanced by dithiol ligands [148, 149]. A very interesting property of thiol-encapsulated gold nanoparticles is their activity in catalytic reactions [150]. From a chemical point of view, the attached thiol molecules should passivate the active gold surface. However, the thiolate encapsulation did not block the catalytic sites in a significant way, as proved by electrooxidation experiments. An explanation for this unusual behavior is the encapsulation of gold oxides formed in the organic layer. The uptake of oxygen and the subsequent expansion of the microstructure of the gold oxides led to an increase in particle size and the formation of nanoporosity in the shell.

3.2. Surface Modification of Metal Oxide Nanoparticles with Monomolecular Layers Figure 8. Silane coupling agents often used in the surface modification of silica particles.

Often the surface atoms of metal oxide nanoparticles carry terminal or bridging O− or OH groups, less often Cl, OR, or carboxylate and other complexing groups. These surface

Core-Shell Nanoparticles

groups are typical reactive sites for the attachment of suitable organic groups. Coordinative bonds are less favorable because the interactions are too weak and because the metal atoms at the surface usually have no pronounced Lewisacidic properties, that is, no empty coordination sites. Many metal oxide nanoparticles contain chemical reactive hydroxy groups on their surface. These can easily react with organically substituted silane molecules of the principal formula Rn SiX4−n , where X is a hydrolyzable group, such as a halogen, alkoxy, or amine group. The rest R is connected by a hydrolytically stable bond, for example, a Si C bond. The OH groups on the surface of the nanoparticles react via hydrolysis and condensation reactions with these so-called silane coupling agents. A plethora of functional groups can be deposited on the particle surface with this approach, and it is beyond the scope of this chapter to mention all possible combinations. Depending on the functional group, specific surface properties can be introduced which are often important for further modifications, for example, to enhance the interfacial interaction between the inorganic particle and a polymer shell. Most of the work done with these compounds was on the modification of silica systems; however, in principal reactions with other nanoparticles that contain hydroxy groups on their surface can also be imagined, such as alumina [151]. The work on organic surface modification of metal oxide nanoparticles by approaches other than those previously mentioned is limited. However, the chemical and structural principles behind the various surface modification methods can be discussed for clusters in which the overall geometry and surface structure are well defined. Because most of the modification methods of clusters can also be transferred to particles, and, based on the fact that there is a fluent transition between clusters and nanoparticles, a clear distinction between these two systems cannot be made. Therefore, we will also present here possibilities of surface modification of clusters with organic molecules to form core-shell-type architectures. In the postsynthesis modification approach, the functionalized cluster or particle is formed in two steps that are distinctly separate from each other: the cluster/particle core is formed first, and the functional organic groups are introduced later in a different reaction. The alternative method is synthesis of the oxo clusters or particles in the presence of functional organic molecules (i.e., the functionalization of the clusters/particles occurs in-situ). The assembly of the cluster occurs in such a way that the organic groups are exclusively bonded to the surface atoms. The advantage of this method is that the process is based on a self-limiting organization of the inorganic and organic building blocks as the formation of the cluster core and the capping of the cluster surface by the organic groups mutually influence each other. The growth of the core is controlled and limited by the organic groups, and, vice versa, the incorporation of the organic groups on the cluster surface is controlled by the chemical reactions by which the cluster is formed. This technique is especially valuable for transition metal oxo systems formed in sol–gel reactions. For example, the surface-modified clusters Zr6 (OH)4 O4 (OMc)12 (OMc = methacrylate) are formed when Zr(On Pr)4 is reacted with

207 4 molar equivalents of methacrylic acid. The molecular structure reveals the existence of chelating and bridging methacrylate ligands which cover the cluster surface [152]. Clusters of different sizes and shapes and a different degree of substitution by organic ligands can be obtained by modifying the reaction conditions. The main parameters appear to be the nature of the metal alkoxide, the metal alkoxide/carboxylic acid ratio, and the type of carboxylate groups [153–161]. The results obtained for isolated and structurally characterized metal oxide clusters are models for the chemistry leading to the functionalization of larger metal oxide particles. An example of the use of bidendate ligands in the synthesis of surface-modified nanoparticles is the hydrolysis of different metal alkoxides in the presence of acetylacetone and a noncomplexing organic acid, which, after aging, led to nanoparticles with tunable sizes in the range of 2 to 4 nm [162–164].

3.3. Polymer Shells The attachment of polymers to the surface of nanoparticles is obtained by the initiation of the polymerization via a surface-immobilized initiator “grafting-from” approach, by the attachment of the polymer via active sites to the particle surface “grafting-to” approach, or by encapsulation techniques in dispersions. The “grafting-from” method has advantages over the “grafting to” method particularly because the latter has limits in the surface coverage due to diffusion restrictions caused by already immobilized macromolecules. Ionic and radical polymerizations were performed from the surfaces of nanoparticles. In general, these polymerization techniques can be divided into controlled and free radical polymerizations. The controlled polymerizations have the advantage that they possess the possibility to grow polymer shells with the desired thickness and composition. In contrast, free radical polymerizations usually show a very fast kinetics with many termination reactions occurring that do not allow a distinct control over the shell morphology.

3.3.1. Grafting-From Approach Several polymerization techniques were used for the grafting-from approach (Table 1). Generally this method allows the highest density of polymer chains on the surface of a particle. For this technique the initiator molecules are usually attached to the particle surface in a first step, and afterward the polymerization is promoted. Ionic Polymerizations Polyesters were grafted from various ultrafine inorganic particles such as silica, titania, and ferrite by anionic ring-opening copolymerization of epoxides and cyclic acid anhydrides [165]. The silica, titania, and ferrite particles had sizes of 16, 120, and 15 nm, respectively, and surface hydroxy group concentrations of 1.37, 0.77, and 0.55 mmol/g. The initiating group was a potassium carboxylate, which was attached to the surface by a series of modifications. The amount of initiating groups on the surface of the silica, titanium oxide, and ferrite particles was 1.92, 0.94, and 0.62 mmol/g, respectively. The organic monomers used were styrene oxide, chloromethyloxirane, gylcidyl methacrylate, gylcidyl phenyl ether, phthalic anhydride, succinic anhydride, and maleic anhydride.

Core-Shell Nanoparticles

208 Table 1. Examples of polymerization reaction initiated from inorganic particles.

Initiator

Inorganic particle

Polymerization method

SiO2 , TiO2 , Ferrite

Anionic Ring Opening

Styrene oxide, chloromethyloxirane, gylcidyl methacrylate, gylcidyl phenyl ether, phthalic anhydride, succinic anhydride, maleic anhydride

[166]

Au

Living cationic

2-Ethyl-2-oxazoline, 2-Phenyl-2-oxazoline

[167]

SiO2

Living anionic

Styrene

[168]

SiO2

Free radical

Styrene

[169]

SiO2

Free radical

MMA

[170]

SiO2

Free radical

MMA

[173]

SiO2

Controlled radical

Styrene

[174]

SiO2 , Au

Controlled radical

N -Butyl acrylate

[176, 178]

Au

Ring-opening metathesis polymerization

Different norbornenyl-containing monomers

[181]

Surface-modified gold nanoparticles were used to initiate the living cationic ring-opening polymerization of 2-ethyl-2oxazoline and 2-phenyl-2-oxazoline [166]. The grafting-from polymerization was initiated by 11-hydroxyundecane-1-thiol attached to the surface, which was activated by triflate. In this manner, dense polymer brushes were prepared in a so-called one-pot multistep reaction. Ex situ kinetic studies of the polymerization of 2-phenyl-2-oxazoline using Fourier transform infrared spectroscopy and matrix-assisted laser desorption ionization time-of-flight mass spectrometry resulted in a linear relation between the reaction time and degree of polymerization of the grafted polymer. The

Monomer

Ref.

polymers were successfully end-functionalized by termination with secondary amines. Living anionic polymerization was initiated from 12– 20-nm 1,1-diphenylethylene (DPE) surface-modified silica nanoparticles [167]. N -BuLi was used to activate the DPE, which allowed the anionic polymerization of styrene in benzene. A high-vacuum reactor was used to allow polymerization from the surface of silica particles under anhydrous solution conditions. Free Radical Polymerizations Diazo groups capable of initiating free radical polymerization were attached to the surface of pyrogenic amorphous silica (Aerosil) by a

Core-Shell Nanoparticles

multistep synthesis [168]. PS was grafted from the thus modified surface. It should be noted that in free radical polymerizations with the described macroinitiators, polymerization also occurs in the solution because the surface-attached initiator decomposes into two radicals from which one initiates polymerization in solution. Therefore, with this technique a purification step is necessary to separate the core-shell systems from the homopolymer. A linear dependence between the monomer and initiator concentration and the polymer coverage was observed. In several studies, silane coupling agents such as alkoxysilanes [169] or chlorosilanes [170] have been used to modify the surface of silica particles. For example, 4,4′ azobis(4-cyanopentanoic acid) was immobilized on the surface of pyrogenic silica via amide bonds through attached aminophenyltrimethoxysilane. The concentration of diazo groups at the surface reached 0.2 mmol/g. DSC measurements proved that the stability of the initiator decreased upon immobilization. Poly(methyl methacrylate) (PMMA) with a molecular weight of up to 8 7 × 105 was grafted from the particle surface. The polymerization was highly affected by the Trommsdorff effect [171], which was also thought to be responsible for the higher molecular weights of the grafted polymer compared with the homopolymer formed in solution. Surface-anchored peroxide initiators were used for the grafting of MMA [172]. This type of initiator was attached to the surface by treatment of silica particles with thionyl chloride to obtain surface Si-Cl groups, which were then reacted with tert-butylhydroperoxide or diisopropylbenzene hydroperoxide to form the initiating groups. Depending on the type of peroxide, the concentration of initiators on the surface was 0.08 mmol/g for tert-butylperoxide and 0.29 mmol/g for diisopropylbenzene peroxide. Both peroxide groups were able to initiate the polymerization and led to the grafting of PMMA. The study of the kinetics and mechanism of an azoinitiated free radical styrene polymerization from modified silica particles was part of more detailed studies for the grafting of polymers from surfaces [170]. In these studies initiators with a cleavable ester group were attached to the surface. Atom Transfer Radical Polymerizations In contrast to free radical polymerizations, in which, polymerization also occurs in the solution, this is not the case for controlled radical polymerization, such as transition metal-mediated atom transfer radical polymerization (ATRP). Dense silica spheres with a diameter of 70 nm were produced with the Stöber process and modified by reaction with the silane coupling agent [2-(4-chloromethylphenyl)ethyl]dimethylethoxysilane [173]. A surface concentration of the initiator of 0.14 mmol/g was thus achieved. The particles obtained were used as macroinitiators for ATRP of styrene. In a similar approach (11′ -chlorodimethylsilylundecyl)-2-chloro2-phenylacetate was covalently attached to the surface of commercially available silica particles, and ATRP grafting of styrene from the surface was achieved. The grafted polymers were detached from the solid particles for analysis. After polymerization of a first generation of grafts

209 and work-up of the hybrid particles, the chain ends of the grafts were still active in initiating a second monomer feed [174]. A soft polymer shell around 12-nm silica particles was formed with the use of a chlorodimethylsilylfunctionalized 2-bromo-2-methylpropionate derivative and n-butylacrylate as the monomer [175]. Hyperbranched polymer-silica hybrid nanoparticles were synthesized by surface-initiated self-condensing ATRP of an acrylic AB∗ inimer from 16-nm silica nanoparticles [176]. About 0.13 mmol initiator/g of silica particles and 0.4 initiator molecules/nm2 were immobilized on the silica particles, corresponding to about 320 initiator molecules per silica particle. 2-(2-Bromopropionyloxy)ethylacrylate and tertbutylacrylate were used as monomers. Gold nanoparticles were modified with thiole-functionalized initiators for ATRP, and n-butylacrylate was polymerized from the surface [177]. Furthermore, CdS/SiO2 core-shell nanoparticles were prepared and modified with an ATRP initiator, followed by the polymerization of MMA [178]. These hybrid inorganic polymer nanoparticles could be cast into films that retained the photoluminescence of the precursor CdS particles and showed an even distribution of the CdS/SiO2 cores throughout the PMMA matrix. An alternative controlled/“living” radical polymerization technique that was used in the grafting of polymers from silica surfaces is the alkoxyamine-based method that also allowed the controlled synthesis of polymer brushes with variable sizes on the surface of the inorganic core [179]. Ring-Opening Metathesis Polymerizations Transitionmetal-catalyzed ring-opening metathesis polymerization (ROMP) was used to graft different norbornenyl-containing monomers from the surface of colloidal gold particles [180]. The two-step preparation of metathesis-ready gold nanoparticles includes the preparation of 1-mercapto-10-(exo-5norbornen-2-oxy)decane and its immobilization on the metal particles. Metathesis of the norbornene rings on the particles with catalyst Cl2 Ru(PCy3 2 CHPh was achieved in less than 10 min in CDCl3 ; subsequent addition of a redox-active ferrocenyl norbornene complex led to copolymerization. ROMP was also carried out from the surface of cadmium selenide nanoparticles stabilized by functional phosphine oxides [181]. These phosphine oxides bear a vinyl group that allows a carbene exchange with the metathesis catalyst and the subsequent polymerization of a variety of cyclic olefins.

3.3.2. Grafting-to Approach The grafting-to approach is based on the reaction of preformed functional polymers containing reactive groups with the surface of inorganic particles. This includes the reaction with the bare surface, for example, in the case of silica with silanol groups at the surface, or the reaction of chemically modified surfaces that have already been treated with silane coupling agents. Either weak (hydrogen bonds, van der Waals) or strong (covalent, coordinative) chemical interactions are feasible. The reaction of colloidal silica of 10-, 45-, and 120-nm diameter with trimethoxysilyl-terminated poly(maleic anhydride-co-styrene) and poly(maleic anhydride-co-MMA) in THF resulted in surface modification. The maximum

210 graft densities of 5 and 3.5 chains/nm2 were obtained in the reaction of the silica colloids of 10- and 45-nm diameter [182]. Citrate-capped gold nanoparticles can be efficiently grafted with a covalently attached polymer monolayer a few nanometers thick, by simple contact of the metal surface with diluted aqueous solutions of hydrophilic polymers that are end-capped with disulfide moieties [183]. The end-functionalized polymers were synthesized with the use of disulfide-functionalized diazo initiators and N -[tris-(hydroxymethyl)methyl]acrylamide as well as N -(isopropryl)acrylamide as monomers. The hydrophilic polymer-coated gold colloids can be freeze-dried and stored as powders that can be subsequently dissolved to yield stable aqueous dispersions, even at very large concentration. Another interesting approach to polymer-capped nanoparticles is the use of end-functionalized polymers, which can interact with the metal species as so-called macroligands, control the final particle size, and stabilize the particles in solution. Nanoscopic CdS entities were produced with this approach, with a thiol-modified poly(caprolactone) as the macroligand [184]. The clusters were grown through the reaction of thiourea with cadmium acetate, and the macroligands were in competition with this reaction by the surface stabilization of the formed particles. There are a variety of other methods that allow the grafting of polymers to particle surfaces; the interested reader is referred to reviews on this topic [185–188].

3.3.3. Polymer Encapsulation One of the major problems in the formation of polymer shells on the surfaces of inorganic particles is the chemical incompatibility of the surface and the monomers (Fig. 9). To overcome this difficulty, coupling agents with functional groups, as specified previously, are allowed to react with the surface functional groups. These groups either participate in the polymerization process or simply change the hydrophobicity of the surface. Another possibility for enhancing the polymerization on the particle surface in a heterophase polymerization is the growth of the macromolecules on the surface by the adsorption of required components, namely, surfactants, monomers, or initiator molecules. Interactions that typically take place with the compounds are charge-charge attractions, hydrogen bonding, or hydrophobic interactions. Polymerization occurs primarily at the surface of unmodified particles because of the adsorption of the monomer on the surface, followed by polymerization in the adsorbed layer [189–191]. The formation of an initial hydrophobic layer on the surface of the particles seems to be crucial for the formation of the polymer shell. Methods for the introduction of such a layer include the use of nonionic surfactants, amphiphilic block copolymers, etc. Furthermore, these surface-active molecules fulfill another task: they avoid agglomeration of the particles. The concentration of such compatibilizing molecules in solution must be carefully tuned because a latex formation in free micelles in the emulsion can be observed at higher concentrations. Moreover, these molecules are usually only weakly bonded to the surface and can therefore desorb easily. Hence, the covalent attachment of organic groups, which

Core-Shell Nanoparticles

Figure 9. Common approaches to core-shell particles. Adapted with permission from [216], J.-E. Jönsson et al., Macromolecules 27, 1932 (1994). © 1994, American Chemical Society.

potentially can interact in the polymerization reaction, is used for surface modification. Silane coupling agents are ideal for this purpose, especially in the case of oxidic inorganic particles containing reactive groups on their surface [192–199]. With a one-pot process Nanosized surfacemodified silica particles encapsulated with a polymeric layer have been formed (Fig. 10). Depending on the reaction conditions like stabilizer concentration, different products were obtained, for example, those in which a few particles were encapsulated in a polymer shell and those where only single-shell particles were formed. Additionally, pure polymer latex particles can also be formed in an emulsion. Spherical CaCO3 -PS composite nanoparticles were obtained by pretreating CaCO3 particles with stearic acid or -methacryloxypropyltrimethoxysilane to make them more hydrophobic and using these precursors in different emulsion polymerizations [200, 201]. Polymerization in microemulsions is probably the most promising versatile technique for the synthesis of a wide variety of nanoparticles with polymeric shells because this

Figure 10. Principle of the formation of 3D networks by site-specific interactions in the shell of nanoparticles.

Core-Shell Nanoparticles

technique offers the possibility of controlling precisely both the size of the core and that of the polymeric shell formed. This is especially important if the properties of the core-shell particles are dependent on both parameters. One example is magnetite particles, which are employed in biology and medicine for the magnetic separation of biochemical products and cells as well as the magnetic guidance of particle systems for site-specific drug delivery. Because the size, charge, and surface chemistry of magnetic particles could strongly influence their biodistribution, it is important to have good control over all of these parameters. In a recent study it was proved that magnetite nanoparticles with a core radius as small as 15 nm and a PMA-poly(hydroxyethyl methacrylate) random copolymer shell size between 80 and 320 nm can be synthesized in a single microemulsion [202]. An interesting unconventional technique used for the formation of organic shells is sonochemistry [203, 204]. This method was used for the quantitative coating of -Fe2 O3 nanoparticles with octadecyltrihydrosilane and the formation of titania nanoparticles with a polyaniline shell. The advantage of this approach is that the aggregation of nanoparticles can be broken down under ultrasonic irradiation, and through the formation of a shell a new aggregation is avoided. Armes et al. reported the synthesis of colloidal silicaconducting polymer core-shell nanocomposites by the oxidative polymerization of either aniline or pyrrole via dispersion polymerization in aqueous media on spherical or “stringy” silica nanoparticles [205–210]. The authors found that both the particle size and the chemical composition depend markedly on the type of chemical oxidant selected for the in-situ polymerization of the monomers. In addition, other functional cores were also used for the surface pyrrole polymerization, such as silica-coated magnetite particles [211] or tin oxide particles [212]. The resulting colloidal poly(pyrrole)-silica-magnetite nanocomposites have a conducting polymer content of 75 mass% and exhibit superparamagnetism. Furthermore, it was shown that a strong acid-base interaction between vinyl monomers such as 2-vinylpyridine, 4-vinylpyridine, and 2-(diethylamino)ethyl methacrylate and the silica surface promotes the formation of nanocomposites in the aqueous media [213]. In addition to studies on silica cores trapped in conducting polymers, other inorganic cores were also used. For example, HAuCl4 was used as an oxidizing agent for the polymerization of pyrrole and the simultaneous formation of Au nanoparticles within the core [214].

4. ORGANIC–ORGANIC CORE-SHELL NANOPARTICLES 4.1. Polymer–Polymer Core-Shell Systems In recent years spherical polymer cross-linked particles (microgels) have attracted attention in the production of micro- and nanostructured materials. In addition to their specific properties, such as the swelling behavior in certain solvents, a variety of functionalizations can be attached to the surface of these particles for further modifications. The major advantage of the microgel systems compared with other colloids is their stabilization by the solution. If the

211 appropriate solvent is used the particles are filled and surrounded by the solvent and therefore are practically dissolved within the molecular medium on a nanoscopic scale. The cross-link density can easily control the volume swelling behavior. One of the most commonly used polymer cores consists of cross-linked PS, but other monomers for latex particles are also in use, which are typically formed by applying free radical emulsion polymerization in combination with selected quantities of cross-linking agents. An overview of different routes towards core-shell polymer nanoparticles is given in Figure 9. The methodologies used are (i) two-stage seeded emulsion polymerization; (ii) emulsion polymerization using reactive surfactants; (iv) stepwise heterocoagulation of smaller particles onto larger ones, followed by heat treatment at a temperature above the Tg of the shells; and (v) the use of reactive block copolymers that allow polymerization [215]. These routes are generally employed for micrometer-sized particles but can also be employed for nanoparticles. Because of the plethora of different systems already investigated, we will limit the description here to some principal modifications of the latex particles. Some recent examples of different core-shell nanoparticles that were prepared by various types of emulsion polymerizations are PMMA-PS [216], polypyrrole-PS [217], PS-poly(p-vinylphenol) [218], poly(dimethylsiloxane)poly(butadiene) [219], and cross-linked PS-poly-(tert-butyl acrylate) [220]. Some of these polymer–polymer core-shell architectures have potential applications for biomaterials. For example, PS-poly(glycidyl methacrylate) core-shell nanoparticles with mean diameters of about 85 nm were modified with a number of different ligands, including diamines of increasing chain length, amino acids and corresponding amines, and higher molecular weight ligands like polymyxin B [221]. The modified particles were tested for their endotoxin (ET) binding capacity in water and physiological sodium chloride solution with the Limulus amebocyte lysate assay. It was found that the ET binding properties of the different surface modifications depend on the ability of the ligand to form Coulomb and van der Waals interactions with the ET molecule influenced by the nature of the suspension medium. Therefore, the choice of ligands for particle modification has to consider minutely the conditions under which ET has to be removed, for example, removal from pure water, dialysis fluids, plasma, or blood. PS latexes were functionalized with halogen-containing groups at their surface capable for initiating ATRP and applied as macroinitiators in this polymerization technique. With this method hydrophobic core-hydrophilic shell nanoparticles were prepared with the use of hydroxyethyl acrylate, 2-(methacryloxy)ethyl trimethylammonium chloride, or methacrylic acid as monomers [222, 223]. A technologically important issue for the application of conducting polymers is to increase their processibility by attaching them to latex particles. Because many of the conducting polymers are water-soluble the surface of the microgel needs at least to be partially polar. To reach this property latex particles were, for example, synthesized from carboxylic or sulfonic derivatized PS. These particles (120 nm)

212 are soluble in aqueous solutions and were used as a seed in a second stage for the synthesis of the core-shell particles. Polypyrrole, for example, can be polymerized around the core in an oxidative chemical in-situ polymerization [224]. In a similar approach styrene-butadiene-methacrylate latex (200 nm) was covered with polyaniline, polypyrrole, and poly(3-methoxythiophene) [225]. Polypyrrole was deposited from aqueous media on a poly(ethylene oxide)-stabilized PS latex with a mean diameter of 129 nm. The polypyrrole loading was varied systematically. The resulting polypyrrole-PS composites did not possess the expected “core-shell” morphology previously observed for other polypyrrole-coated PS particles. Instead, TEM studies indicated that the conducting polymer component was present as discrete 20–30-nm nanoparticles, which were adsorbed to the PS latex [217]. An example of core-crosslinked micelles is the use of amphiphilic poly(ethylene glycol)-b-polylactide (PEG/PLA) copolymers with an aldehyde group at one chain end and a methacryloyl group at the other terminal end. The methacryloyl groups located in the core of the micelle were polymerized to form core-shell-type nanoparticles with diameters between 20 and 30 nm and reactive aldehyde groups on the surface [226]. Microgels composed of environmentally responsive polymers continue to attract attention because of their potential applications in numerous fields, including drug delivery, chemical separations, catalysis, and sensors. Perhaps the most studied class of responsive polymers is thermoresponsive poly(alkylacrylamides). Hence, these polymers were also used in the synthesis of stimuli-responsive core-shell latexes [227]. Many of the synthesized systems contain a responsive and a nonresponsive component. Furthermore, multiresponsive systems were also investigated that are composed of two or more environmentally sensitive polymers [228]. For example, core particles composed of cross-linked poly(N -isopropylacrylamide) or mixtures of the latter with acrylic acid were synthesized via precipitation polymerization and then used as nuclei for subsequent polymerization for the formation of a shell. These hydrogel particles display both a strong temperature and pH dependence on swelling. Through a change of the composition of the core and the shell, for example, by copolymerizing other monomers, the response of these hydrogels is made tunable [229]. As already mentioned, the sol–gel process is a versatile tool if an oxidic shell should be formed around a particle. The advantage of this process is its applicability for a lot of different metals (e.g., Si, Ti, Zr, Sn, etc.) and the mild reaction conditions that allow the incorporation of organic groups into the networks formed. An interesting aspect of these coated polymer particles is the removal of their core by calcination and the production of hollow spheres.

4.2. Cross-Linked Micelles Amphiphilic block copolymers can aggregate in solutions into spherical micelles depending on the solvents and concentrations between solvent and block copolymers. If the copolymers contain functional groups that are able to undergo cross-linking reactions, stable spherical core-shell nanostructures can be formed with either a cross-linked core

Core-Shell Nanoparticles

or shell and with differing physical and chemical properties of the core and the shell that can be tailored over a wide range. Wooley et al. prepared several examples of so-called shell cross-linked knedel-like (SCK) polymer assemblies with diameters between 5 and 100 nm [230]. Based on the crosslinked shell, these nanostructures have a robust character and surface stability under changing environmental conditions. In addition, the shell serves as a membrane that has a permeability that can be tailored to control the transport of guests to and from the core. Furthermore, the lack of crosslinks in the core region maintains chain mobility and access to the core volume. Hence these systems are nanoscale scaffolds that can mimic several biological systems. One important parameter that influences the properties of these systems is the composition and structure of the amphiphilic block copolymer. Another parameter is the composition and cross-link density of the shell that determines the extent of swelling, controls the interactions with the environment, and influences the shell’s permeability. The core provides a unique nanoenvironment for guests that are able to pass the shell barrier. Typical examples of such systems comprise assemblies of PS-poly(4-vinylpyridine) block copolymers in which the shell was cross-linked by applying quaternization reactions of the pyridine segments [231] or PS-poly(acrylic acid) cross-linked via amidation reactions in the shell [232]. Other reports used micelles of PS-poly(2cinnamoylethyl methacrylate) or polyisoprene-poly(2cinnamoylethyl methacrylate) diblock copolymers with photo-cross-linked shells [233], poly(solketal methacrylate)block-poly(2-(dimethylamino)ethyl methacrylate) with 1, 2-bis(2-iodoethoxyl)ethane cross-linked shells [234], 1, 2bis(2-iodoethoxy)ethane cross-linked partially quaternized 2-(dimethylamino)ethyl methacrylate-N -(morpholino)ethyl methacrylate diblock copolymer micelles [235], or zwitterionic diblock copolymers [236].

5. POLYMER–INORGANIC CORE-SHELL NANOPARTICLES Because of the simple availability of polymer latexes in combination with a wide range of sizes and functionalities, they are ideal building blocks for complex functional materials. For the deposition of precursors for inorganic shells the latex particles often require a charge. This charge can be introduced by applying ionic initiators in the free radical polymerizations. With this technique stable colloidal coreshell particles consisting of a PS core and a titania coating were prepared in one step by the hydrolysis of a titanium alkoxide in the presence of a cationic PS latex. The coatings obtained were very smooth and uniform and varied in thickness from just a few nanometers up to 50 nm [237]. The surface modification with magnetic compounds attracted much interest because of their possible applications for therapeutic or analytical purposes. In the first case, magnetic particle carriers permit either the guiding and release of a drug in a specific site of the body or the extraction of tumor cells from the organism and their curing in vitro [238, 239]. For this purpose different latex particles were covered with magnetic substances. For example, PSpoly(N -isopropylacrylamide) core-shell nanoparticles were

Core-Shell Nanoparticles

prepared, and iron oxide was adsorbed on the outer shell via electrostatic interactions [240]. Gold shells on PS latex were prepared by adsorbing polyethyleneimine polyelectrolyte to the polymer core via electrostatic and hydrogen-bonding attraction, followed by the adsorption of negatively charged Au particles with sizes of 3 nm or 15 nm [241]. The coverage of gold was increased by a seeding procedure, resulting in a Au shell thickness of 10 ± 5 nm. The coating of anionic or cationic latexes with metals or metal salts is a general approach to core-shell nanoparticles. Examples of such systems are PS latexes with shells of Y basic carbonate [242], Zr compounds [243], TiO2 [244], iron compounds [245], or copper and copper compounds [246]. These particles were synthesized via absorption and reaction of metal salts on the surface of the latexes. In several cases these particles were calcined and hollow spheres were obtained. A different method for producing multilayer particles, for example, of a latex core with a corona of inorganic nanoparticles capping its surface, is the layer-by-layer method. Magnetite, titanium dioxide, silica, and laponite nanoparticles were used as the inorganic building blocks for multilayer formation on PS sphere templates [247, 248]. Composite organic–inorganic particles were formed by the controlled assembly of the preformed nanoparticles in alternation with oppositely charged polyelectrolytes onto PS microspheres. Inorganic nanoparticles with different diameters were shown to work effectively in this coating process. In addition, hollow capsules can be obtained from the precursor polymer– inorganic core-shell particles by direct removal of the core material by chemical or physical methods.

6. INORGANIC BIOMOLECULE CORE-SHELL SYSTEMS In recent years nature motivated intense efforts to develop assembly methods that mimic or exploit the recognition capabilities and interactions found in biological systems. Many studies concentrated on fundamental principles by applying surface-modified nanoparticles with complementary receptor-substrate sites on their surface for the detection of biological molecules and/or for the recognition-driven self-assembly of ordered aggregates. The basic idea for the latter approach was transferred from biological systems in which organic molecules show a remarkable level of control over the nucleation and mineral phase of inorganic materials such as calcium carbonate and silica, and over the assembly of crystallites and other nanoscale building blocks into complex structures required for biological functions. Of particular value are methods that could be applied to materials with interesting electronic or optical properties. Similar to the interactions in enzymes, organic ligands can interact with the surface of metal or semiconductor nanoparticles. With this approach peptides with limited selectivity for binding to metal and semiconductor surfaces can, for example, be used for the particle modification [249].

213 A general approach is to prepare nanoparticles by wet chemistry in the presence of stabilizing agents that coordinate to the surface, such as citrates, phosphines, or thiols, of the resulting nanoparticles. These molecules prevent an uncontrolled growth and aggregation and are substituted in a further step with ligands that have a stronger interaction with the metal atoms at the surface. For example, gold nanoparticles were reacted with thiol-modified DNA sequences that allow recognition by specific particles of each other through the hydrogen bonds typical for these biomolecules. This technique in particular was used as a method for assembling colloidal nanoparticles rationally and reversibly into macroscopic aggregates [250, 251]. The method involves attachment of two batches of noncomplementary DNA oligonucleotides capped with thiol groups as linkers to the surface of nanometer gold particles. When oligonucleotide duplexes with complementary ends are added to the two grafted sequences, the nanoparticles self-assemble into aggregates (Fig. 10). The assembly process can be reversed by thermal denaturation. This material synthesis approach has been extended to a wide range of biomolecules, including peptides and proteins [252–254], and a collection of nanoparticles, including gold and semiconductor quantum dots [255, 256]. However, gold proved to be ideal for this kind of materials since, for example, it was more difficult to modify CdSe and CdS quantum dots because of a surfactant layer that is very strongly bound to their surfaces and, consequently, difficult to displace [257]. Hence, if nanoparticle compositions different from Au were required, a typical route is to cover their surface with a gold layer and immobilize the oligonucleotides on this layer afterward [258]. Another possibility for activating the surface of quantum dots for the coupling of biomolecules is their coverage with silane coupling agents, which was used in the capping of core(CdSe)-shell(CdS) nanoparticles [107]. Other biorelated interaction techniques, for example, antibodyantigen recognition, were also applied for the directed selfassembly of Au and Ag nanoparticles [259]. The absorption of the biomolecules in these cases was carried out at a pH above the isoelectric point of the stabilizing citrate groups. The proteins then show a strong interaction with colloids by the positively charged amino acid side chains. In addition to the possibility of enhancing the selfassembly of larger arrays, the use of recognition chemistry has found wide application in the development of highly sensitive and selective diagnostic methods [107, 260–264]. Major problems in the synthesis of biomolecule-capped nanoparticles are the avoidance of harsh reaction conditions, which often lead to the inactivation of the biofunctionalization and the strong fixation of the molecules to the surface of the particles. Only a few examples from the field of nanoparticlebiomolecule interactions were given in this chapter. The interested reader is referred to a comprehensive review that shows the possibilities of combining inorganic materials with biomolecules [265]. As a summary of the different materials described within this chapter, we have collected in Table 2 most of the combinations, as well as their properties of interest and corresponding references.

Core-Shell Nanoparticles

214 Table 2. Summary of material combinations and properties. Core

Shell

Synthesis method

Relevant property

Ref.

CdSe CdSe CdSe CdSe CdSe CdSe CdSe CdS CdS InAs InAs Au Ag Au Au Au Ag Ag Ag Ag Au Pt Pd Au PS SiO2 SiO2 SiO2 Fe Co Fe Fe Fe Co Ag

ZnS ZnS ZnSe ZnSe CdS CdS CdS Ag2 S CdSe InP CdSe Ag Au Cd Pb Sn Pb Cd In Hg Pt Au Pt Pd Ag Au Ag Ag Ag Ag Au In Nd Pt Co

Microemulsions TOPO Single-molecule precursors TOPO TOPO Single-molecule precursors Microemulsions Microemulsions Microemulsions TOPO TOPO Chemical reduction Chemical reduction -Radiolysis -Radiolysis -Radiolysis -Radiolysis -Radiolysis -Radiolysis -Radiolysis Chemical reduction -Radiolysis Ethanol reduction

[10] [10] [26] [11, 22, 26] [12, 25] [25] [20, 23] [19] [20] [27] [27] [31, 33, 34, 38] [33, 36] [40] [40, 41] [42] [43, 44] [45] [43] [46] [48] [48] [53, 54] [55] [56] [57–62] [63] [39] [75] [75] [76, 77] [78] [78] [79] [80]

Fe2 O3 Fe3 O4 Fe Au Ag Ag Ag CdS CdSe Ag Au Au Au Fe2 O3 Pd CeO2 Ag

SiO2 SiO2 SiO2 SiO2 SiO2 SiO2 SiO2 SiO2 SiO2 TiO2 TiO2 TiO2 SnO2 CdS Surfactant Surfactant Surfactant PS, PMMA

Chemical reduction Chemical reduction Chemical reduction Electroless plating Microemulsions Microemulsions Microemulsions Solvated metal-atom dispersion Solvated metal-atom dispersion Redox transmetalation Thermal decomposition + transmetalation Sol–gel Sol–gel Sol–gel Sol–gel Sol–gel Microemulsions DMF reduction Sol–gel Sol–gel DMF reduction DMF reduction Layer-by-layer Sol–gel Chemical precipitation Microemulsion Microemulsion Microemulsion Emulsion

Luminescence Luminescence Luminescence Luminescence Luminescence Luminescence Luminescence Luminescence Luminescence IR luminescence IR luminescence Optical absorption Optical absorption Optical absorption Optical absorption Optical absorption Optical absorption Optical absorption Optical absorption Optical absorption Optical absorption Optical absorption Catalysis Catalysis Optical absorption Optical absorption Optical absorption Optical absorption Magnetic Magnetic Magnetic Magnetic Magnetic Magnetic Magnetic Magnetic Magnetic Magnetic Optical absorption Optical absorption Optical absorption Optical absorption Luminescence Luminescence Optical absorption Biocompatible membranes Optical absorption Capacitance Optical absorption Nonlinear optical response

[82, 84] [72, 85, 87–89] [92] [93–96, 99–101, 104, 105] [35, 97, 98, 102, 103] [110] [111] [106, 109] [107, 108] [112] [113] [114] [116] [117] [133, 135, 136] [134] [137] [138]

SiO2 SiO2−n Rn Au Au Pt CdS

3-Aminopropyltriethoxysilane PS p-Mercaptophenol -Substituted alkanethiolates Alkyl isocyanides Hexanedithiol

Stöber process Dispersion/hydrosilation In-situ reduction In-situ reduction/ligand exchange Coordination Coordination

Functionalization with biomolecules

Film formation

[139] [141] [142] [143] [148] [149] continued

Core-Shell Nanoparticles

215

Table 2. Continued Core

Shell

Synthesis method

Al2 O3 Zr, Ti, Nb Metal oxo clusters ZrO2 TiO2 SiO2 , TiO2 , Ni-Zn ferrite Au SiO2

PS, polyacrylamide Carboxylic acids

Dispersion Coordination

[151] [152–161]

Acetylacetone Acetylacetone Polyesters

Coordination Coordination Grafting from polymerization

[162] [163, 164] [165]

Polyoxazoline PS

Grafting from polymerization Grafting from polymerization

SiO2 SiO2 SiO2 Au CdS/SiO2 Au CdSe SiO2

PMMA Poly(n-butyl acrylate) PAA, poly(t-butyl acrylate) Poly(n-butyl acrylate) PMMA Polynorbornene Polycyclooctene Poly(maleic anhydride-co-styrene), Poly(MA-co-Me methacrylate) Poly(N -[tris-(hydroxymethyl) methyl]-acrylamide), poly(N -(isopropryl)acrylamide) Poly(caprolactone) Poly(ethyl acrylate) PS Poly(t-butyl acrylate) PS Cross-linked copolymer of methacrylic acid and hydroxyethyl methacrylate Octadecyltrihydrosilane Polyaniline Polyaniline Polypyrrole Polypyrrole Polypyrrole Poly(vinylpyridine) and copolymers with styrene and MMA, poly(2(diethylamino)ethyl methacrylate) Polypyrrole

Grafting Grafting Grafting Grafting Grafting Grafting Grafting Grafting

[166] [167, 168, 170, 173, 174, 179] [169, 172, 179] [175] [176] [177] [178] [180] [181] [182]

Au

CdS SiO2 SiO2 SiO2 CaCO3 Fe3 O4

Fe2 O3 TiO2 SiO2 SiO2 SiO2 /magnetite SnO2 SiO2

Au PMMA PS PS Poly (dimethylsiloxane) PS Poly (t-butyl acrylate) PS PS

PS Thermoplastic rubber latexes Poly (ethylene glycol)

PS Polypyrrole Poly(p-vinylphenol) Poly(butadiene) Poly(t-butyl acrylate) PS Poly(glycidyl methacrylate) Poly(2-hydroxyethyl acrylate) poly(2-(methacryloyloxy)ethyl trimethylammonium chloride) PMA, PMMA Polyaniline polypyrrole poly(3-methoxythiophene) Polylactide

from from from from from from from to

Relevant property

polymerization polymerization polymerization polymerization polymerization polymerization polymerization

Ref.

Grafting to

[183]

Grafting to Encapsulation Encapsulation Dispersion Emulsion Microemulsion

[184] [192–195] [196–198] [199] [200, 201] [202]

Sonochemistry Sonochemistry Surface polymerization Surface polymerization Surface polymerization Surface polymerization Surface polymerization

Magnetic

Magnetic Conductivity Conductivity Conductivity Superparamagnetism Conductivity

[203] [204] [205, 206, 210] [207–209] [211] [212] [213]

Polymerization in diblock copolymer micelles Emulsion Emulsion Emulsion Emulsion

[216] [217, 224] [218] [219]

Emulsion

[220]

Emulsion Emulsion

[214]

Endotoxin binding

[221] [222]

Grafting from Emulsion

[223] [225]

Cross-linked micelle

[226] continued

Core-Shell Nanoparticles

216 Table 2. Continued Core

Shell

Synthesis method

Relevant property

Ref.

PS Poly (N -isopropylacrylamide) poly (NIPAm-co-acrylic acid) Poly (N -isopropylacrylamide)

Poly(N -isopropylacrylamide) Poly (N -isopropulacrylamide) poly(NIPAm-co-acrylic acid) Poly(N -isopropylacrylamide) copolymerized with butyl methacrylate Poly(4-vinylpyridine) Poly(acrylic acid) Poly(2-cinnamoyethyl methacrylate) Poly(2-(dimethylamino)ethyl methacrylate) Poly(N -(morpholino)ethyl) methacrylate Poly(2-tetrahydropyranyl methacrylate) TiO2 Iron oxide

Emulsion Precipitation polymerization

Stimuli responsive Stimuli responsive

[227] [228]

Precipitation polymerization

Thermoresponsive

[229]

PS PS Polyisoprene Poly (solketal methacrylate) Poly (2-dimethylamino) ethyl methacrylate Poly (2-dimethylamino) ethyl methacrylate PS PS-poly(Nisopropylacrylamide) PS PS PS PS PS PS PS PS Au ZnS@CdSe Au@Ag

Au Yttrium basic carbonate Zirconium compounds TiO2 Iron compounds Copper and copper compounds Fe3 O4 TiO2 , SiO2 , laponite DNA oligonucleotides Proteins and other biomolecules Oligonucleotides

Shell-cross-linked micelle Shell-cross-linked micelle Shell-cross-linked micelle

[231] [232] [233]

Shell-cross-linked micelle

[234]

Shell-cross-linked micelle

[235]

Shell-cross-linked micelle

[236]

Sol–gel Surface deposition Surface Surface Surface Surface Surface Surface

deposition deposition deposition deposition deposition deposition

Layer-by-layer deposition Layer-by-layer deposition Coordination Electrostatic binding Electrostatic binding

7. CONCLUSIONS In this chapter we have reviewed the synthesis and properties of core-shell nanoparticles of various composition. A classification based on the nature of the core and the shell lead us to distinguish mainly between organic and inorganic materials, but also to find clear subdivisions taking into account the metallic, semiconductor, or dielectric nature of the inorganic materials, or the relative configuration of organic and inorganic material within the composite particles. We have shown that there is a plethora of possible combinations leading to nanomaterials with various functionalities, which has a great potential for applications in technology and life sciences.

GLOSSARY Colloid A substance consisting of very tiny particles that are usually between 1 nm and 1000 nm in diameter and that are suspended in a continuous medium, such as a liquid, a solid, or a gaseous substance. Core-shell nanoparticles Particles with size below 1 m, composed of an inner, typically spherical core homogeneously surrounded by a shell of a different material. Emulsion Drop-shaped distribution of one liquid in another while both of them are not miscible.

Magnetic

[237] [240] [241] [242] [243] [244] [245] [246]

Magnetic

Ultrasensitive biological detection

[247] [248] [250, 251, 254] [253, 256] [258]

Latex A homogeneous colloidal dispersion of polymeric particles in water. Nanoshells Spherical colloid particles with a hollow interior, usually filled with solvent. Plasmon resonance Collective oscillation of conduction electrons in metal nanoparticles, induced by an electromagnetic wave. Quantum dot Particle of matter so small that quantum size effects can be observed. Silica coating Deposition of an amorphous silicon dioxide shell on a preformed colloid particle. Surfactant Abbreviation for surface-active agent; soluble compound that reduces the surface tension of liquids, or reduces interfacial tension between two liquids or a liquid and a solid.

REFERENCES 1. 2. 3. 4. 5. 6. 7.

C.-J. Zhong and M. M. Maye, Adv. Mater. 13, 1507 (2001). E. Bourgeat-Lami, J. Nanosci. Nanotechnol. 2, 1 (2002). G. Kickelbick, Prog. Polym. Sci., 28, 83 (2003). A. P. Alivisatos, J. Phys. Chem. 100, 13226 (1996). A. P. Alivisatos, Science 271, 933 (1996). L. L. Beecroft and C. K. Ober, Chem. Mater. 9, 1302 (1997). N. Herron and D. L. Thorn, Adv. Mater. 10, 1173 (1998).

Core-Shell Nanoparticles 8. U. Woggon, “Optical Properties of Semiconductor Quantum Dots.” Springer-Verlag, Berlin, 1997. 9. T. Vossmeyer, S. Jia, E. Delonno, M. R. Diehl, X. Peng, A. P. Alivisatos, and J. R. Heath, J. Appl. Phys. 84, 3664 (1998). 10. B. O. Dabbousi, J. Rodriguez-Viejo, F. V. Mikulec, J. R. Heine, H. Mattoussi, R. Ober, K. F. Jensen, and M. G. Bawendi, J. Phys. Chem. B 101, 9463 (1997). 11. M. Danek, K. F. Jensen, C. B. Murray, and M. G. Bawendi, Chem. Mater. 8, 173 (1996). 12. X. Peng, M. C. Schlamp, A. V. Kadavanich, and A. P. Alivisatos, J. Am. Chem. Soc. 119, 7019 (1997). 13. M. A. Hines and P. Guyot-Sionnest, J. Phys. Chem. 100, 468 (1996). 14. A. Eychmüller, A. Hasselbarth, and H. Weller, J. Lumin. 53, 113 (1992). 15. A. Hasselbarth, A. Eychmüller, M. Eichberger, M. Giersig, A. Mews, and H. Weller, J. Phys. Chem. 97, 5333 (1993). 16. V. F. Kamalov, R. Little, S. L. Logunov, and M. A. El-Sayed, J. Phys. Chem. 100, 6381 (1996). 17. H. S. Zhou, H. Sasahara, I. Homma, H. Komiyama, and J. Haus, Chem. Mater. 6, 1534 (1994). 18. G. Counio, S. Esnouf, T. Gacoin, and J. P. Boilot, J. Phys. Chem. 100, 20021 (1996). 19. M. Y. Han, W. Huang, C. H. Chew, L. M. Gan, X. J. Zhang, and W. Ji, J. Phys. Chem. B 102, 1884 (1998). 20. Y. Tian, T. Newton, N. A. Kotov, D. M. Guldi, and J. H. Fendler, J. Phys. Chem. 100, 8927 (1996). 21. D. Diaz, J. Robles, T. Ni, S.-E. Castillo-Blum, D. Nagesha, O.-J. A. Fregoso, and N. A. Kotov, J. Phys. Chem. B 103, 9854 (1999). 22. C. F. Hoener, K. A. Allan, A. J. Bard, A. Campion, A. M. Fox, T. E. Mallouk, S. E. Webber, and J. M. White, J. Phys. Chem. 96, 3812 (1992). 23. E. Hao, H. Sun, Z. Zhou, J. Liu, B. Yang, and J. Shen, Chem. Mater. 11, 3096 (1999). 24. C. B. Murray, D. J. Norris, and M. G. Bawendi, J. Am. Chem. Soc. 115, 8706 (1993). 25. N. Revaprasadu, M. A. Malik, P. O’Brien, and G. Wakefield, Chem. Commun. 1573 (1999). 26. M. A. Malik, P. O’Brien, and N. Revaprasadu, Chem. Mater. 14, 2004 (2002). 27. Y.-W. Cao and U. Banin, Angew. Chem., Int. Ed. 38, 3692 (1999). 28. Y.-W. Cao and U. Banin, J. Am. Chem. Soc. 122, 9692 (2000). 29. A. Henglein, J. Phys. Chem. 97, 54 (1993). 30. P. Mulvaney, Langmuir 12, 788 (1996). 31. R. H. Morriss and L. F. Collins, J. Chem. Phys. 41, 3357 (1964). 32. J. Sinzig, U. Radtke, M. Quinten, and U. Kreibig, Z. Phys. D 26, 242 (1993). 33. L. Rivas, S. Sánchez-Cortes, J. V. García-Ramos, and G. Morcillo, Langmuir 16, 9722 (2000). 34. L. Lu, H. Wang, Y. Zhou, S. Xi, H. Zhang, J. Hu, and B. Zhao, Chem. Commun. 144 (2002). 35. T. Ung, L. M. Liz-Marzán, and P. Mulvaney, Langmuir 14, 3740 (1998). 36. Y. W. Cao, R. Jin, and C. A. Mirkin, J. Am. Chem. Soc. 123, 7961 (2001). 37. Y. Sun, B. T. Mayers, and Y. Xia, NanoLetters 2, 481 (2002). 38. M. Schierhorn and L. M. Liz-Marzán, NanoLetters 2, 13 (2002). 39. Y. Kobayashi, V. Salgueiriño-Maceira, and L. M. Liz-Marzán, Chem. Mater. 13, 1630 (2001). 40. F. Henglein, A. Henglein, and P. Mulvaney, Ber. Bunsenges. Phys. Chem. 98, 180 (1994). 41. P. Mulvaney, M. Giersig, and A. Henglein, J. Phys. Chem. 97, 7061 (1993). 42. A. Henglein and M. Giersig, J. Phys. Chem. 98, 6931 (1994). 43. A. Henglein, P. Mulvaney, A. Holzwarth, T. E. Sosebee, and A. Fojtik, Ber. Bunsenges. Phys. Chem. 96, 754 (1992).

217 44. A. Henglein, A. Holzwarth, and P. Mulvaney, J. Phys. Chem. 96, 8700 (1992). 45. A. Henglein, P. Mulvaney, and A. Holzwarth, J. Phys. Chem. 96, 2411 (1992). 46. L. Katsikas, M. Gutiérrez, and A. Henglein, J. Phys. Chem. 100, 11203 (1996). 47. A. Henglein and M. Giersig, J. Phys. Chem. B 104, 5056 (2000). 48. A. Henglein, J. Phys. Chem. B 104, 2201 (2000). 49. A. Henglein, J. Phys. Chem. B 104, 6683 (2000). 50. N. Toshima, in “Fine Particle. Synthesis, Characterization, and Mechanism of Growth” (T. Sugimoto, Ed.), pp. 430–459. Dekker, New York, 2000. 51. N. Toshima and T. Yonezawa, New J. Chem. 22, 1179 (1998). 52. T. Yonezawa and N. Toshima, J. Chem. Soc., Faraday Trans. 91, 4111 (1995). 53. N. Toshima, K. Kushihashi, T. Yonezawa, and H. Hirai, Chem. Lett. 1769 (1989). 54. N. Toshima, Y. Shiraishi, A. Shiotsuki, D. Ikenaga, and Y. Wang, Eur. Phys. J. D 16, 209 (2001). 55. A. F. Lee, C. J. Baddeley, C. Hardacre, R. M. Ormerod, R. M. Lambert, G. Schmid, and H. West, J. Phys. Chem. 99, 6096 (1995). 56. P. Barnickel and A. Wokaun, Mol. Phys. 67, 1355 (1989). 57. S. L. Westcott, S. J. Oldenburg, L. Randall, and N. J. Halas, Langmuir 14, 5396 (1998). 58. S. J. Oldenburg, R. D. Averitt, S. L. Westcott, and N. J. Halas, Chem. Phys. Lett. 288, 243 (1998). 59. G. Mie, Ann. Phys. 25, 337 (1908). 60. S. J. Oldenburg, J. B. Jackson, S. L. Westcott, and N. J. Halas, Appl. Phys. Lett. 75, 2897 (1999). 61. S. J. Oldenburg, S. L. Westcott, R. D. Averitt, and N. J. Halas, J. Chem. Phys. 111, 4729 (1999). 62. C. Graf and A. van Blaaderen, Langmuir 18, 524 (2002). 63. J. B. Jackson and N. J. Halas, J. Phys. Chem. B 105, 2743 (2001). 64. G. O. Mallory and J. B. Hajdu, Eds., “Electroless Plating: Fundamentals and Applications.” American Electroplaters and Surface Finishers Society, Orlando, FL, 1990. 65. V. K. Pecharsky and K. A. Gschneidner, Jr., J. Magn. Magn. Mater. 200, 44 (1999). 66. C. Kormann, E. Schwab, F.-W. Raulfs, and K. H. Beck, U.S. Patent 5,500,141, 1996. 67. J. C. Lodder, D. J. Monsma, R. Vlutters, and T. Shimatsu, J. Magn. Magn. Mater. 198–199, 119 (1999). 68. A. Jordan, R. Scholz, P. Wust, H. Schirra, T. Schiestel, H. Schmidt, and R. Felix, J. Magn. Magn. Mater. 194, 185 (1999). 69. M. Y. Mapra, I. J. Körner, M. Hildebrandt, R. Bargou, D. Krahl, P. Reichardt, and B. Dörken, Blood 89, 337 (1997). 70. S. A. Iakovenko, A. S. Trifonov, M. Giersig, A. Mamedov, D. K. Nagesha, V. V. Hanin, E. C. Soldatov, and N. A. Kotov, Adv. Mater. 11, 388 (1999). 71. T. Fried, G. Shemer, and G. Markovich, Adv. Mater. 13, 1158 (2001). 72. F. G. Aliev, M. A. Correa-Duarte, A. Mamedov, J. W. Ostrander, M. Giersig, L. M. Liz-Marzán, and N. A. Kotov, Adv. Mater. 11, 1006 (1999). 73. M. Giersig and M. Hilgendorff, J. Phys. D: Appl. Phys. 32, L111 (1999). 74. R. D. Shull and L. H. Bennet, Nanostruct. Mater. 1, 83 (1992). 75. J. Rivas, R. D. Sánchez, A. Fondado, C. Izco, A. J. García-Bastida, J. García-Otero, J. Mira, D. Baldomir, A. González, I. Lado, M. A. López-Quintela, and S. B. Oseroff, J. Appl. Phys. 76, 6564 (1994). 76. C. T. Seip and C. J. O’Connor, Nanostruct. Mater. 12, 183 (1999). 77. E. E. Carpenter, C. Sangregorio, and C. J. O’Connor, IEEE Trans. Magn. 35, 3496 (1999). 78. D. Zhang, G. Glavee, K. J. Klabunde, G. C. Hadjipanayis, and C. M. Sorensen, High Temp. Mater. Sci. 36, 93 (1997). 79. J.-I. Park and J. Cheon, J. Am. Chem. Soc. 123, 5743 (2001).

218 80. N. S. Sobal, M. Hilgendorff, H. Moehwald, M. Giersig, M. Spasova, T. Radetic, and M. Farle, NanoLetters 2, 621 (2002). 81. R. K. Iler, Patent 2,885,366, 1959. 82. M. Ohmori and E. Matijevic, J. Colloid Interface Sci. 150, 594 (1992). 83. W. Stöber, A. Fink, and E. Bohn, J. Colloid Interface Sci. 26, 62 (1968). 84. D. M. Thies-Weesie and A. P. Philipse, Langmuir 11, 4180 (1995). 85. A. P. Philipse, M. P. van Bruggen, and C. Pathmamanoharan, Langmuir 10, 92 (1994). 86. A. P. Philipse, A. M. Nechifor, and C. Pathmamanoharan, Langmuir 10, 4451 (1994). 87. Q. Liu, Z. Xu, J. A. Finch, and R. Egerton, Chem. Mater. 10, 3938 (1998). 88. M. A. Correa-Duarte, M. Giersig, N. A. Kotov, and L. M. LizMarzán, Langmuir 14, 6430 (1998). 89. L. N. Donselaar and A. P. Philipse, J. Colloid Interface Sci. 212, 14 (1999). 90. C. Graf, W. Schärtl, K. Fischer, N. Hugenberg, and M. Schmidt, Langmuir 15, 6170 (1999). 91. C. Roos, M. Schmidt, J. Ebenhoch, F. Baumann, B. Deubzer, and J. Weis, Adv. Mater. 11, 761 (1999). 92. M. Ohmori and E. Matijevic, J. Colloid Interface Sci. 160, 288 (1993). 93. L. M. Liz-Marzán and A. P. Philipse, J. Colloid Interface Sci. 176, 459 (1995). 94. L. M. Liz-Marzán, M. Giersig, and P. Mulvaney, J. Chem. Soc., Chem. Commun. 731 (1996). 95. L. M. Liz-Marzán, M. Giersig, and P. Mulvaney, Langmuir 12, 4329 (1996). 96. L. M. Liz-Marzán and P. Mulvaney, New J. Chem. 1285 (1998). 97. M. Giersig, L. M. Liz-Marzán, T. Ung, D. Su, and P. Mulvaney, Ber. Bunsenges. Phys. Chem. 101, 1617 (1997). 98. M. Giersig, T. Ung, L. M. Liz-Marzán, and P. Mulvaney, Adv. Mater. 9, 570 (1997). 99. S. S. Chang, C. W. Shih, C. D. Chen, W. C. Lai, and C. R. Wang, Langmuir 15, 701 (1999). 100. S. O. Obare, N. R. Jana, and C. J. Murphy, NanoLetters 1, 601 (2001). 101. S. R. Hall, S. A. Davis, and S. Mann, Langmuir 16, 1454 (2000). 102. V. V. Hardikar and E. Matijevic, J. Colloid Interface Sci. 221, 133 (2000). 103. Y. Yin, Y. Lu, Y. Sun, and Y. Xia, NanoLetters 2, 427 (2002). 104. E. Mine, A. Yamada, Y. Kobayashi, M. Konno, and L. M. LizMarzán, manuscript submitted for publication. 105. R. I. Nooney, T. Dhanasekaran, Y. Chen, R. Josephs, and A. E. Ostafin, Adv. Mater. 14, 529 (2002). 106. M. A. Correa-Duarte, M. Giersig, and L. M. Liz-Marzán, Chem. Phys. Lett. 286, 497 (1998). 107. M. Bruchez, Jr., M. Moronne, P. Gin, S. Weiss, and A. P. Alivisatos, Science 281, 2013 (1998). 108. D. Gerion, F. Pinaud, S. C. Williams, W. J. Parak, D. Zanchet, S. Weiss, and A. P. Alivisatos, J. Phys. Chem. B 105, 8861 (2001). 109. A. L. Rogach, D. K. Nagesha, J. W. Ostrander, M. Giersig, and N. A. Kotov, Chem. Mater. 12, 2676 (2000). 110. T. Li, J. Moon, A. A. Morrone, J. J. Mecholsky, D. R. Talhman, and J. H. Adair, Langmuir 15, 4328 (1999). 111. I. Pastoriza-Santos and L. M. Liz-Marzán, Langmuir 15, 948 (1999). 112. I. Pastoriza-Santos, D. Koktysch, A. Mamedov, N. A. Kotov, and L. M. Liz-Marzán, Langmuir 16, 2731 (2000). 113. D. S. Koktysh, X. Liang, B.-G. Yun, I. Pastoriza-Santos, R. L. Matts, M. Giersig, C. Serra-Rodríguez, L. M. Liz-Marzán, and N. A. Kotov, Adv. Funct. Mater. 12, 255 (2002). 114. K. S. Mayya, D. I. Gittins, and F. Caruso, Chem. Mater. 13, 3833 (2001).

Core-Shell Nanoparticles 115. K. S. Mayya, D. I. Gittins, A. M. Dibaj, and F. Caruso, NanoLetters 1, 727 (2001). 116. G. Oldfield, T. Ung, and P. Mulvaney, Adv. Mater. 12, 1519 (2000). 117. P. V. Kamat and B. Shanghavi, J. Phys. Chem. B 101, 7675 (1997). 118. R. Nayak, J. Galsworthy, P. Dobson, and J. Hutchinson, J. Mater. Res. 13, 905 (1998). 119. T. Hirano, T. Oku, and K. Suganuma, J. Mater. Chem. 9, 855 (1999). 120. V. Pillai and D. O. Shah, Surfactant Sci. Ser. (Ind. App. Microemulsions), 66, 227 (1997). 121. F. J. Arriagada and K. Osseo-Asare, Colloids Surf. 69, 105 (1992). 122. F. J. Arriagada and K. Osseo-Assare, J. Colloid Interface Sci. 170, 8 (1995). 123. J. Esquena, T. F. Tadros, K. Kostarelos, and C. Solans, Langmuir 13, 6400 (1997). 124. F. J. Arriagada and K. Osseo-Asare, J. Colloid Interface Sci. 211, 210 (1999). 125. P. D. Moran, J. R. Bartlett, J. L. Woolfrey, G. A. Bowmaker, and R. P. Cooney, Ceram. Trans. (Ceram. Processing Sci. Technol.) 51, 27 (1995). 126. V. Chhabra, V. Pillai, B. K. Mishra, A. Morrone, and D. O. Shah, Langmuir 11, 3307 (1995). 127. G. L. Li and G. H. Wang, Nanostruct. Mater. 11, 663 (1999). 128. M. Wu, J. Long, A. Huang, Y. Luo, S. Feng, and R. Xu, Langmuir 15, 8822 (1999). 129. H. Schmidt, C. Kropf, T. Schiestel, H. Schirra, S. Sepeur, and C. Lesniak, Ceram. Trans. 95, 49 (1998). 130. M. L. Curri, G. Palazzo, G. Colafemmina, M. Della Monica, and A. Ceglie, Prog. Colloid Polym. Sci. (Trends Colloid Interface Sci. XII) 110, 188 (1998). 131. P. A. Dresco, V. S. Zaitsev, R. J. Gambino, and B. Chu, Langmuir 15, 1945 (1999); L. Liz, M. A. López-Quintela, J. Rivas, and J. Mira, J. Mater. Sci. 29, 3797 (1994). 132. H. Herrig and R. Hempelmann, Mater. Lett. 27, 287 (1996). 133. L. Guo, Z. Wu, T. Liu, and S. Yang, Physica E 8, 199 (2000). 134. V. Arcoleo, M. Goffredi, A. Longo, and V. Turco Liveri, Mater. Sci. Eng., C 6, 7 (1998). 135. T. Liu, L. Guo, Y. Tao, Y. B. Wang, and W. D. Wang, Nanostruct. Mater. 11, 487 (1999). 136. T. Liu, L. Guo, Y. Tao, T. D. Hu, Y. N. Xie, and J. Zhang, Nanostruct. Mater. 11, 1329 (2000). 137. Z. Wu, R. E. Benfield, L. Guo, H. Li, Q. Yang, D. Grandjean, Q. Li, and H. Zhu, J. Phys.: Condens. Matter 13, 5269 (2001). 138. L. Quaroni and G. Chumanov, J. Am. Chem. Soc. 121, 10642 (1999). 139. A. Van Blaaderen and A. Vrij, J. Colloid Interface Sci. 156, 1 (1993). 140. W. Schärtl, Adv. Mater. 12, 1899 (2000). 141. W. Schärtl, G. Lindenblatt, A. Strack, P. Dziezok, and M. Schmidt, Prog. Colloid Polym. Sci. 110, 285 (1998). 142. M. Brust, J. Fink, D. Bethell, D. J. Schiffrin, and C. Kiely, J. Chem. Soc., Chem. Commun. 1655 (1995). 143. M. J. Hostetler, S. J. Green, J. J. Stockes, and R. W. Murray, J. Am. Chem. Soc. 118, 4212 (1996). 144. C. J. Kiely, J. Fink, M. Brust, D. Bethell, and D. J. Schiffrin, Nature 396, 444 (1998). 145. C. J. Kiely, J. Fink, J. G. Zheng, M. Brust, D. Bethell, and D. J. Schiffrin, Adv. Mater. 12, 640 (2000). 146. M. Sastry, in “Handbook of Surfaces and Interfaces of Materials” (H. S. Nalwa, Ed.), Vol. 3, p. 87. Academic Press, San Diego, 2001. 147. M. Brust and C. J. Kiely, Colloids Surf., A 202, 175 (2002). 148. S. L. Horswell, C. J. Kiely, I. A. O’Neil, and D. J. Schiffrin, J. Am. Chem. Soc. 121, 5573 (1999). 149. K. Hu Kai, M. Brust, and A. J. Bard, Chem. Mater. 10, 1160 (1998). 150. C.-J. Zhong and M. M. Maye, Adv. Mater. 13, 1507 (2001). 151. M. Z. Rong, L. Q. Ji, M. Q. Zhang, and K. Friedrich, Eur. Polym. J. 38, 1573 (2002). 152. G. Kickelbick and U. Schubert, Chem. Ber. 130, 473 (1997).

Core-Shell Nanoparticles 153. G. Kickelbick and U. Schubert, Eur. J. Inorg. Chem. 159 (1998). 154. G. Kickelbick, P. Wiede, and U. Schubert, Inorg. Chim. Acta 284, 1 (1999). 155. G. Kickelbick and U. Schubert, J. Chem. Soc., Dalton Trans. 1301 (1999). 156. B. Moraru, S. Gross, G. Kickelbick, G. Trimmel, and U. Schubert, Monatsh. Chem. 132, 993 (2001). 157. U. Schubert, E. Arpac, W. Glaubitt, A. Helmerich, and C. Chau, Chem. Mater. 4, 291 (1992). 158. B. Moraru, N. Hüsing, G. Kickelbick, U. Schubert, P. Fratzl, and H. Peterlik, Chem. Mater. 14, 2732 (2002). 159. G. Kickelbick and U. Schubert, Eur. J. Inorg. Chem. 159 (1998). 160. B. Moraru, G. Kickelbick, and U. Schubert, Eur. J. Inorg. Chem. 1295 (2001). 161. L. G. Hubert-Pfalzgraf, V. Abada, S. Halut, and J. Roziere, Polyhedron 16, 581 (1997). 162. M. Chatry, M. In, M. Henry, C. Sanchez, and J. Livage, J. Sol-Gel Sci. Technol. 2, 233 (1994). 163. E. Scolan and C. Sanchez, Chem. Mater. 10, 3217 (1998). 164. E. Scolan, C. Magenent, D. Massiot, and C. Sanchez, J. Mater. Chem. 9, 2467 (1999). 165. N. Tsubokawa, A. Kogure, and Y. Sone, J. Polym. Sci.: Part A: Polym. Chem. 28, 1923 (1990). 166. R. Jordan, N. West, A. Ulman, Y.-M. Chou, and O. Nuyken, Macromolecules 34, 1606 (2001). 167. Q. Zhou, S. Wang, X. Fan, R. Advincula, and J. Mays, Langmuir 18, 3324 (2002). 168. R. Laible and K. Hamann, Angew. Makromol. Chem. 48, 97 (1975). 169. G. Boven, M. L. C. M. Oosterling, G. Chella, and A. J. Schouten, Polymer 31, 2377 (1990). 170. O. Prucker and J. Rühe, Macromolecules 31, 602 (1998). 171. G. V. Schulz and G. Harborth, Makromol. Chem. 1, 106 (1948). 172. N. Tsubokawa and H. Ishida, J. Polym. Sci.: Part A: Polym. Chem. 30, 2241 (1992). 173. T. von Werne and T. E. Patten, J. Am. Chem. Soc. 121, 7409 (1999). 174. H. Böttcher, M. L. Hallensleben, S. Nuß, and H. Wurm, Polym. Bull. 44, 223 (2000). 175. G. Carrot, S. Diamanti, M. Manuszak, B. Charleux, and J.-P. Vairon, J. Polym. Sci., Part A: Polym. Chem. 39, 4294 (2001). 176. H. Mori, D. C. Seng, M. Zhang, and A. H. E. Müller, Langmuir 18, 3682 (2002). 177. S. Nuß, H. Bottcher, H. Wurm, and M. L. Hallensleben, Angew. Chem., Int. Ed. 40, 4016 (2001). 178. S. C. Farmer and T. E. Patten, Chem. Mater. 13, 3920 (2001). 179. M. Husseman, E. E. Malmström, M. McNamara, M. Mate, D. Mecerreyes, D. G. Benoit, J. L. Hedrick, P. Mansky, E. Huang, T. P. Russell, and C. J. Hawker, Macromolecules 32, 1424 (1999). 180. K. J. Watson, J. Zhu, S. T. Nguyen, and C. A. Mirkin, J. Am. Chem. Soc. 121, 462 (1999). 181. H. Skaff, M. F. Ilker, E. B. Coughlin, and T. Emrick, J. Am. Chem. Soc. 124, 5729 (2002). 182. K. Yoshinaga, Y. Tani, and Y. Tanaka, Colloid Polym. Sci. 280, 85 (2002). 183. C. Mangeney, F. Ferrage, I. Aujard, V. Marchi-Artzner, L. Jullien, O. Ouari, E. D. Rekaie, A. Laschewsky, I. Vikholm, and J. W. Sadowski, J. Am. Chem. Soc. 124, 5811 (2002). 184. G. Carrot, S. M. Scholz, C. J. G. Plummer, and J. G. Hilborn, Chem. Mater. 11, 3571 (1999). 185. R. Laible and K. Hamann, Adv. Colloid Interface Sci. 13, 65 (1980). 186. K. Ryan, Chem. Ind. 359 (1988). 187. B. Vincent, Chem. Eng. Sci. 48, 429 (1993). 188. K. Yoshinaga, Surf. Sci. Ser. (Fine Particles) 92, 626 (2000). 189. W. D. Hergeth, M. Peller, and P. Hauptmann, Acta Polym. 37, 468 (1986). 190. W. D. Hergeth, P. Starre, and K. Schmutzler, Polymer 29, 1323 (1988).

219 191. W. D. Hergeth, U. J. Steinau, H. J. Bittrich, G. Simon, and K. Schmutzler, Polymer 30, 254 (1989). 192. E. Bourgeat-Lami, P. Espiard, A. Guyot, S. Briat, C. Gauthier, G. Vigier, and J. Perez, ACS Symp. Ser. (Hybrid Organic–Inorganic Composites) 585, 112 (1995). 193. E. Bourgeat-Lami, P. Espiard, and A. Guyot, Polymer 36, 4385 (1995). 194. P. Espiard and A. Guyot, Polymer 36, 4391 (1995). 195. E. Bourgeat-Lami, P. Espiard, A. Guyot, C. Gauthier, L. David, and G. Vigier, Angew. Makromol. Chem. 242, 105 (1996). 196. E. Bourgeat-Lami and J. Lang, J. Colloid Interface Sci. 197, 293 (1998). 197. E. Bourgeat-Lami and J. Lang, J. Colloid Interface Sci. 210, 281 (1999). 198. E. Bourgeat-Lami and J. Lang, Macromol. Symp. 151, 377 (2000). 199. I. Sondi, T. H. Fedynyshyn, R. Sinta, and E. Matijevi, Langmuir 16, 9031 (2000). 200. N. Bechthold, F. Tiarks, M. Willert, K. Landfester, and M. Antonietti, Macromol. Symp. 151, 549 (2000). 201. J. Yu, J. Yu, Z.-X. Guo, and Y.-F. Gao, Macromol. Rapid Commun. 22, 1261 (2001). 202. P. A. Dresco, V. S. Zaitsev, R. J. Gambino, and B. Chu, Langmuir 15, 1945 (1999). 203. K. V. P. M. Shafi, A. Ulman, A. Dyal, X. Yan, N.-L. Yang, C. Estournes, L. Fournes, A. Wattiaux, H. White, and M. Rafailovich, Chem. Mater. 14, 1778 (2002). 204. H. Xia and Q. Wang, Chem. Mater. 14, 2158 (2002). 205. M. Gill, Michael, J. Mykytiuk, S. P. Armes, J. L. Edwards, T. Yeates, P. J. Moreland, and C. Mollett, J. Chem. Soc., Chem. Commun. 108 (1992). 206. M. Gill, F. L. Baines, and S. P. Armes, Synth. Met. 55, 1029 (1993). 207. S. Maeda and S. P. Armes, J. Colloid Interface Sci. 159, 257 (1993). 208. S. Maeda and S. P. Armes, J. Mater. Chem. 4, 935 (1994). 209. R. Flitton, J. Johal, S. Maeda, and S. P. Armes, J. Colloid Interface Sci. 173, (1995). 210. J. Stejskal, P. Kratochvil, S. P. Armes, S. F. Lascelles, A. Riede, M. Helmstedt, J. Prokes, and I. Krivka, Macromolecules 29, 6814 (1996). 211. M. D. Butterworth, S. P. Armes, and A. W. Simpson, J. Chem. Soc., Chem. Commun. 2129 (1994). 212. S. Maeda and S. P. Armes, Synth. Met. 69, 499 (1995). 213. C. Barthet, A. J. Hickey, D. B. Cairns, and S. P. Armes, Adv. Mater. 11, 408 (1999). 214. S. T. Selvan, J. P. Spatz, H.-A. Klok, and M. Möller, Adv. Mater. 10, 132 (1998). 215. W.-H. Li and H. D. H. Stöver, Macromolecules 33, 4354 (2000). 216. J.-E. Jönsson, H. Hassander, and B. Törnell, Macromolecules 27, 1932 (1994). 217. D. B. Cairns, S. P. Armes, and L. G. B. Bremer, Langmuir 15, 8052 (1999). 218. C. Lu and R. Pelton, Colloids Surf., A 201, 161 (2002). 219. M. Okaniwa, Polymer 41, 453 (1999). 220. S. Kirsch, A. Doerk, E. Bartsch, H. Sillescu, K. Landfester, H. W. Spiess, and W. Maechtle, Macromolecules 32, 4508 (1999). 221. R. Darkow, T. Groth, W. Albrecht, K. Lützow, and D. Paul, Biomaterials 20, 1277 (1999). 222. M. M. Guerrini, B. Charleux, and J.-P. Vairon, Macromol. Rapid Commun. 21, 669 (2000). 223. K. Min, J. Hu, C. Wang, and A. Elaissari, J. Polym. Sci.: Part A: Polym. Chem. 40, 892 (2002). 224. A. Yassar, J. Roncali, and F. Garnier, Polym. Commun. 28, 103 (1987).

220 225. C. F. Liu, T. Maruyama, and T. Yamamoto, Polym. J. 25, 363 (1993). 226. M. Iijima, Y. Nagasaki, T. Okada, M. Kato, and K. Kataoka, Macromolecules 32, 1140 (1999). 227. C. Pichot, A. Elaissari, D. Duracher, F. Meinier, and F. Sauzedde, Macromol. Symp. 175, 285 (2001). 228. C. D. Jones and L. A. Lyon, Macromolecules 33, 8301 (2000) and references therein. 229. D. Gan and L. A. Lyon, J. Am. Chem. Soc. 123, 7511 (2001). 230. K. L. Wooley, J. Polym. Sci.: Part A: Polym. Chem. 38, 1397 (2000). 231. K. B. Thurmond II, T. Kowalewski, and K. L. Wooley, J. Am. Chem. Soc. 118, 7239 (1996). 232. H. Huang, T. Kowalewski, E. E. Remsen, R. Gertzmann, and K. L. Wooley, J. Am. Chem. Soc. 119, 11653 (1997). 233. J. Tao, G. Liu, J. Ding, and M. Yang, Macromolecules 30, 4084 (1997). 234. Z. Zhang, G. Liu, and S. Bell, Macromolecules 33, 7877 (2000). 235. V. Buetuen, N. C. Billingham, and S. P. Armes, J. Am. Chem. Soc. 120, 12135 (1998). 236. V. Buetuen, A. B. Lowe, N. C. Billingham, and S. P. Armes, J. Am. Chem. Soc. 121, 4288 (1999). 237. A. Imhof, Langmuir 17, 3579 (2001). 238. R. Langer, Science 249, 1527 (1990). 239. T. Lea, F. Vartdal, K. Nustad, S. Funderud, A. Berge, T. Ellingsen, R. Schmid, P. Stenstad, and J. Ugelstad, J. Mol. Recognit. 1, 9 (1988). 240. F. Sauzedde, A. Elaissari, and C. Pichot, Colloid Polym. Sci. 277, 846 (1999). 241. T. Ji, V. G. Lirtsman, Y. Avny, and D. Davidov, Adv. Mater. 13, 1253 (2001). 242. N. Kawahashi and E. Matijevic, J. Colloid Interface Sci. 138, 534 (1990). 243. N. Kawahashi, C. Persson, and E. Matijevic, J. Mater. Chem. 1, 577 (1991). 244. H. Shiho and N. Kawahashi, Colloid Polym. Sci. 278, 270 (2000). 245. H. Shiho and N. Kawahashi, J. Colloid Interface Sci. 226, 91 (2000). 246. N. Kawahashi and H. Shiho, J. Mater. Chem. 10, 2294 (2000).

Core-Shell Nanoparticles 247. F. Caruso, M. Spasova, A. Susha, M. Giersig, and R. A. Caruso, Chem. Mater. 13, 109 (2001). 248. R. A. Caruso, A. Susha, and F. Caruso, Chem. Mater. 13, 400 (2001). 249. S. R. Whaley, D. S. English, E. L. Hu, P. F. Barbara, and A. M. Belcher, Nature 405, 665 (2000). 250. C. A. Mirkin, R. L. Letsinger, R. C. Mucic, and J. J. Storhoff, Nature 382, 607 (1996). 251. S.-J. Park, A. A. Lazarides, C. A. Mirkin, and R. L. Letsinger, Angew. Chem., Int. Ed. 40, 2909 (2001). 252. S. Mann, W. Shenton, M. Li, S. Connolly, and D. Fitzmaurice, Adv. Mater. 12, 147 (2000). 253. H. Mattoussi, J. M. Mauro, E. R. Goldman, G. P. Anderson, V. C. Sundar, F. V. Mikulec, and M. G. Bawendi, J. Am. Chem. Soc. 122, 12142 (2000). 254. A. P. Alivisatos, K. P. Johnsson, X. Peng, T. E. Wilson, C. J. Loweth, M. P. Bruchez, Jr., and P. G. Schultz, Nature 382, 609 (1996). 255. D. Gerion, W. J. Parak, S. C. Williams, D. Zanchet, C. M. Micheel, and A. P. Alivisatos, J. Am. Chem. Soc. 124, 7070 (2002). 256. W. C. W. Chan and S. Nie, Science 281, 2016 (1998). 257. G. P. Mitchell, C. A. Mirkin, and R. L. Letsinger, J. Am. Chem. Soc. 121, 8122 (1999). 258. Y. Cao, R. Jin, and C. A. Mirkin, J. Am. Chem. Soc. 123, 7961 (2001). 259. W. Shenton, S. A. Davis, and S. Mann, Adv. Mater. 11, 449 (1999). 260. T. A. Taton, R. L. Letsinger, and R. C. Mirkin, Science 289, 1757 (2000). 261. T. A. Taton, G. Lu, and C. A. Mirkin, J. Am. Chem. Soc. 123, 5164 (2001). 262. J. J. Storhoff, R. Elghanian, R. C. Mucic, C. A. Mirkin, and R. L. Letsinger, J. Am. Chem. Soc. 120, 1959 (1998). 263. Y. C. Cao, R. Jin, and C. A. Mirkin, Science 297, 1536 (2002). 264. J.-M. Nam, S.-J. Park, and C. A. Mirkin, J. Am. Chem. Soc. 124, 3820 (2002). 265. C. M. Niemeyer, Angew. Chem., Int. Ed. 40, 4128 (2001).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Crystal Engineering of Metallic Nanoparticles A. Hernández Creus, Y. Gimeno Universidad de La Laguna, La Laguna, Spain

R. C. Salvarezza, A. J. Arvia Universidad Nacional de La Plata—CONICET, La Plata, Argentina

CONTENTS 1. Introduction 2. Basic Electrochemical Aspects 3. Electrochemical Preparation 4. Crystal Growth Kinetics 5. Applications 6. Conclusions Glossary References

1. INTRODUCTION Metallic nano/mesoparticles supported on conducting substrates such as carbons are of interest due to their unique physical chemistry properties, making them of importance in different fields of applied science [1–10] and for the eventual development of new techniques. The technology used to produce preformed mass selected atomic clusters on carbons involves nanometer-sized systems in which, as a result of the reduced average atomic coordination, the quantum size effect, and the modified screening response, clusters display novel electronic and magnetic properties [11–12]. The physicochemical properties of nano/mesoparticles depend strongly on the perturbations that arise from the large area/volume atom ratio in relation to bulk metals. Therefore, the manageability of pattern morphology of nano/mesoislands is of importance in dealing with catalysts, electrocatalysis, and nano/microfabrication and in developing materials with interesting properties such as giant magneto resistances, high saturation-moments, and high anisotropic binary clusters. A great research effort has been made to develop a predictive understanding of the relationship between particle size and structure, and

ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

physicochemical properties such as electronic and magnetic properties. The understanding of the reactivity of supported nanometer-sized particles in heterogeneous catalysis can be approached by either a molecular standpoint in which small soft-landed metal clusters of 5–50 atoms are considered, or by taking into account the reactivity of extended singlecrystal surfaces. Both approaches have advantages and limitations, particularly to explain the behavior of supported clusters a few nm in diameter, due to their intrinsic heterogeneities [11–12]. Significant progress in this direction has been made in the understanding of size-dependent catalytic behavior, showing that the high activity requires the minimum size compatible with metallic properties. In contrast, the fabrication of the nano/mesoparticles remains at present a rather empirical task due to the fact that the physical processes that determine shape and size at the nanometer scale are not systematically identified. Metal nano/mesoparticles can be preparated by different techniques, such as vapor deposition, ultrahigh vapor deposition, chemical vapor deposition, chemical impregnation, homogeneous reactions in solution, electroless deposition, and electrodeposition. The success in obtaining particles with defined physicochemical properties relies on the possibility of obtaining a narrow particle size distribution. Precisely, major efforts in nano/microparticle preparation by electrochemical methods have focused on improving particle size distribution. At present, very homogeneous size particles can be prepared by nucleation and growth from solution [1–10], although this goal is still a pending issue for obtaining metallic particles supported on solid substrates [13–14]. For the latter, most of the present methods result in the formation of particles with a broad particle size distribution function, the broadening being dependent on growth conditions.

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (221–235)

222 In principle, metal electrodeposition appears as an excellent route for nano/mesoparticle preparation on conducting substrates due to the easy control of the experimental variables [13–18]. In this method the reaction rate can be easily adjusted either for a surface reaction or for a reaction under a mass-transport kinetic regime. Therefore, metal electrodeposition can be taken as a model system to investigate possible relationships between particle size and shape, and growth kinetics. Recently, the preparation of narrowly dispersed particles with average diameter at the nano/micro scale and relative standard deviation RSD ≤ 30% by electrodeposition has been accomplished [13–14]. In this case, mesoparticles have been modeled as simple hemispherical objects produced either by using double pulse electrodeposition to nucleate first a large number of particles that later are allowed to grow slowly on the substrate, or metal and hydrogen codeposition under forced convection due to bubbles at the microscale [13–14]. For further advances in the above mentioned matters, one should understand the entire kinetics and mechanisms of the process at different length and time scales. This fundamental knowledge would provide a rational approach to prepare mesoparticles with different structures and shapes, to discover how physical and chemical processes determine particle structure and shape, and to determine the dependence of the physicochemical properties of mesoparticles on their structure and shape. Manipulation of mesoparticles is a “nanostructuring” problem, deserving theoretical and experimental consideration. Therefore, a combination of theoretical models that have been developed to understand the relationship between growth kinetics, shape, and structure based on both Euclidean and fractal concepts, and powerful experimental techniques able to image these objects at the nanometer scale, such as scanning tunneling microscopy (STM) and atomic force microscopy (AFM), is required. In this contribution, the state-of-the-art concerning our ability to handle physical variables of remarkable influence in determining the structure/shape of metal nano/mesoparticles produced by electrodeposition on highly oriented pyrolytic graphite (HOPG), that is, the basal plane of graphite (C(0001)), and to evaluate their catalytic activity for some test reactions is reviewed. Among the different carbon substrates HOPG exhibits unique properties for studies at the nanometer level by scanning probe microscopies (STM and AFM) because it is atomically smooth and chemically inert. Metal electrodeposition on HOPG was taken as a model system to disperse metallic particles, because it is a smooth substrate in which no metal underpotential deposition takes place, as in no complete metallic adlayer is produced [19]. In fact, due to the large difference between the metal and the carbon surface tension, metal particle deposition on HOPG follows the Volmer– Weber mechanism [20] rather than the Stranski–Krastanov [21] and the van der Merwe mechanisms [22] (Fig. 1). Furthermore, metallic particles dispersed on carbon are important in dealing with a number of electrocatalytic and heterogeneous catalytic systems [23]. Metal electrodeposition on HOPG has also been taken as a model system for the development of a new technology for nanostructure fabrication1 at wafer scale [24] and for the preparation of

Crystal Engineering of Metallic Nanoparticles

Figure 1. Different growth modes of metal deposition on foreign substrate. (a) Volmer–Weber mechanism, 3-D metallic islands. (b) Frank–van der Merwe mechanism, layer by layer growth. (c) Stranski–Krastanov mechanism, 3-D islands on top of 2-D layer.

metallic wires [25, 26] as promising interconnectors in future nanoscale electronics. Metal-patterned HOPG has also been considered for directly embossing different polymeric materials [27]. The presentation of the material is organized as follows. First, some fundamental aspects of electrochemistry related to metal electrodeposition are described. Then, the preparation methods of metal nano/mesoparticles are considered. Special attention to the electrodeposition of crystals of platinum, palladium, and gold on HOPG is paid. These are typical examples of metal electrodeposition from which kinetic conclusions applicable to the growth of other metallic phases can be drawn. The correlation between kinetics and particle structure from the early stages of growth onward is described. Later, the electrodeposition parameters and the size and shape of HOPG-supported crystals are related to the physicochemical processes that determine the shape and structure at different temporal and length scales. In the third section, we compare nanoparticle growth by electrochemical techniques with vapor deposition methods. The influence of the applied electric potential on crystal shape is discussed. Finally, typical applications of metallic deposits supported on HOPG in nanowires, embossing, and electrocatalysis are given.

2. BASIC ELECTROCHEMICAL ASPECTS 2.1. Substrate Characterization HOPG is commonly used in STM and AFM studies [28] because it exhibits micrometer-sized atomically smooth terraces separated by steps 0.35 nm in height (Fig. 2a) that

223

Crystal Engineering of Metallic Nanoparticles

Figure 2. STM images of freshly exfoliated HOPG surface. (a) 2 × 2 m2 , large and smooth terraces separated by steps are clearly seen. (b) Atomic resolution, 6 × 6 nm2 . Bright spots correspond to carbon atoms.

correspond to the interlayer spacing. Also, HOPG is chemically inert and can be handled in the atmosphere with neglegible contamination. It is used for x y piezoelectric calibration at the atomic level in STM. The STM imaging of a HOPG surface originates reproducible atomically resolved hexagonal lattices with d = 0246 nm, where three of the six carbon atoms of the honeycomb structure of graphite are imaged (Fig. 2b). The smoothness of the HOPG surface in relation to other carbon substrates such as mechanically polished vitrous carbon, fractured vitrous carbon, and amorphous carbon turns it into the best substrate to study the morphology of nano/mesometallic particles.

2.2. Electrochemical Deposition of Metals Metal particles dispersed on a HOPG surface can most frequently be obtained by electrodeposition from aqueous electrolyte solutions containing a relatively small concentration of depositing ions as either cations or anions. The electrochemical cell (Fig. 3) usually consists of a working cathode, such as a freshly exfoliated HOPG surface, a large area noble metal counter electrode, and a reference electrode.

Figure 3. Conventional three-electrode water-jacket electrochemical cell. (a) Reference electrode. (b) Counter electrode. (c) Working electrode. (d) Inert gas bubbler.

In general, the plating solution is deareated and kept under an inert atmosphere (argon or nitrogen) to suppress any interference of the oxygen electroreduction reaction at the cathode. The electrochemical cell is operated by a potentiostat that permits one to select either a constant potential or a constant current regime under nonstationary conditions when the electrolysis time (td ) is properly adjusted to avoid both the coalescence of growing particles and the formation of a continuous metal layer. Electrodeposition under potential control can be made by applying either a constant potential (Ed ) or a linear potential sweep (Ed = Eo − vtd , where Eo is the starting potential and v is the potential sweep rate). In the former case, the current transient (Ic versus td , where Ic is the electrodeposition current) is recorded. The integration of this transient yielding the electrodeposition charge (Qc ) passes between t = 0 and t = td . For the latter, the electrodeposition voltammogram is obtained, that is, the E versus Ic plot at the rate v, and the value of Qc can also be determined from the voltammogram. For electrodeposition under a constant current (galvanostatic), the potential transient (Ec versus td ) is recorded. This procedure is interesting for electrodeposition under a constant flow of discharging ions on the cathode. The values of Ic and Qc are usually referred to as the unit surface area (A) of the substrate as current density (jc  and charge density (qc ) because a reliable definition of A for dispersed particles on a solid surface is not always feasible. Further details on electrochemical deposition techniques are given in reference [29].

3. ELECTROCHEMICAL PREPARATION 3.1. Previous Work Metal particles of silver, gold, palladium, molybdenum, copper, and platinum can be easily electrodeposited on HOPG [13–18, 25, 26, 30–33]. Platinum mesoparticles (ca. 0.1 m in diameter) on glassy carbon have been electrodeposited from H2 PtCl6 in the potential of the first electroreduction wave (0.2 to 0.5 V versus SCE). At loading levels 10–20 g cm−2 , these electrodes exhibit high particle stability and an exchange current density value for hydrogen evolution comparable to that for bulk platinum [34, 35]. Platinum particles of 2–2.5 nm diameter can also be formed by spontaneous reduction of H2 PtCl6 at the plane edges and basal plane of HOPG caused by the incompletely oxidized surface functional carbon group [36]. Potentiostatic deposition on pyrolytic graphite of roughly circular platinum particles 0.3–1 m in diameter, spatially distributed in a random fashion, has also been reported [37]. The electrodeposition of platinum on HOPG has been followed by AFM tapping mode. In the absence of a supporting electrolyte, two-dimensional (2-D) nanoparticles aggregate after reaching a critical mobile particle size. Subsequent growth is dominated by surface diffusion of mobile platinum particles. In aqueous perchloric acid, three-dimensional (3-D) growth is the dominant mechanism competing with 2-D/3-D growth. This leads to various morphologies. The process is also influenced by the presence of specifically adsorbed anions [38].

224

Crystal Engineering of Metallic Nanoparticles

Platinum nanoparticles have been electrocrystallized on HOPG by a first reactant electrostatic adsorption, followed by electrochemical reduction on a 4-aminophenyl monolayer-grafted carbon substrate. The STM imaging shows a good size monodispersivity and particles well separated from one another [39]. Metal nano- and microparticles of silver, gold, platinum, molybdenum, and nickel narrowly dispersed by electrodeposition on HOPG have been formed by a two-step method—a 5 msec potential step to nucleate particles at an overvoltage c = −05 V, and a growth pulse set in the range −002 ≤ c ≤ 0250 V to reach the desired particle diameter. The slow growth process leads to a particle dispersion ranging from 50 nm to 2 m in diameter with a RSD as low 7% [13–14]. Carbon electrodes with an effective fiber radius of 1 nm can be prepared by different procedures: electrochemical etching of a carbon fiber, deposition of electrophoretic paint, and an inverted deposition technique for insulating the tip. Thus, the complete insulation of the whole body of the carbon fiber, except tip pinhole formation, is avoided [40].

3.2. Potentiodynamic Electrodeposition A typical potentiodynamic polarization curve (jc versus Ed ) for metal electrodeposition on HOPG is shown in Figure 4 [41]. The values of Ed are referred to the standard calomel electrode scale. This example corresponds to palladium electrodeposition at v = 5 × 10−3 V/s from a palladium chloride-containing aqueous solution at 298 K [41]. The electrochemical deposition of a metal (Me) can be produced either from the discharge metal of ions (Mez+ ) of charge z, Mez+ aq + ze− = Me0

(1a)

or from the discharge of a metal-containing complex anion (MeXz− n ) containing n ligands and charge z, ′

MeXnz− aq + z′ e− = Me0 + nXz+z −

(1b)

50

-2

jc (µA cm )

0

1

-50

2

-100

4 -150 -0.4

Er

3 0.0

0.4

0.8

Ed (V vs SCE) Figure 4. Potentiodynamic polarization curve of HOPG in aqueous 75 × 10−4 M palladium chloride + 5 × 10−2 M sodium perchlorate + 5 × 10−3 M perchloric acid run from 0.85 V to −0.35 V at v = 5 × 10−3 V/s; 298 K. The cathodic current density (jc ) is referred to the working electrode geometric area. (1) Beginning of metal electrodeposition. (2) Main cathodic current peak. (3) Cathodic limiting current. (4) Hydrogen evolution reaction. The reversible potential (Er  for Pd electrodeposition is indicated.

The reversible potential Er of Equations (1a) and (1b) are given by the following Nernst equations, respectively, Er 1a = E o 1a + RT /zFln aMez+ /aMe0

(2a)

and Er 1b = E o 1b + RT /z′ Fln aMeXz− n  ′

/aMe0aXz+z − n

(2b)

E o N , N = (1a), (1b), and a[i] being the standard potential and the activity of species i, and z, F and T the number of electrons per reacting species involved in the reaction, the Faraday constant (96,500 C/equivalent) and the absolute temperature, respectively. The activity of the metal is taken as 1. For values of Ed more negative than Er , the cathodic potentiodynamic polarization curve first exhibits a low current portion (Fig. 4, region 1) extending from Er to a value of Ed where a rapid increase in jc leading to a current maximum (Fig. 4, region 2) is observed. As Ed is further increased negatively, the cathodic current density attains a limiting value jL (Fig. 4, region 3) that extends to the potential range where the hydrogen electrode reactions on electrodeposited palladium commence (Fig. 4, region 4). Often, the first low current region approaches a Ed versus ln jc linear relation indicating that the kinetics of metal electrodeposition is under charge-transfer control. However, as Ed is shifted more negatively, the kinetics of the electrodeposition reaction turns into a mass transport regime from the solution side to the cathode surface.

3.3. Potentiostatic Electrodeposition Detailed information of the nucleation and growth of metallic particles on foreign substrates can be obtained from potentiostatic current transients. In this case, for diluted solutions, these transients exhibit a rapid increase in current that is frequently observed for the nucleation and growth of metal particles under mass transport control. Thus, following the potential step at Ed , the diffusion zones built up around growing nuclei begin to overlap, and after a time (tM ) the growth kinetics become a linear diffusion-controlled process. Finally, for a tc that depends on Ed , the contribution of free convection increases, and jc approaches the jL value (Fig. 4, region 3). Potentiostatic current transients run at different Ed can be analyzed with growth models that are applicable to the electrochemical formation of a new solid phase on a conducting substrate. Further details related to this subject are extensively described in the literature (see, e.g., [19]). Current transients for palladium electrodeposition on HOPG were obtained by stepping the potential from E ≥ Er to values of Ed lying within the different potential windows of the palladium electrodeposition regions (Fig. 4) [41]. For the potential regions 2–4 (Fig. 4), the current transients (Fig. 5a) exhibit an initial fast current decay due to the contribution of the double-layer charging that is significant for td < 1 sec. Subsequently, a fast increase in current up to a maximum value jM at time tM is observed. Subsequently, a slow decay characterized by a jc versus td−1/2 dependence is observed (Fig. 5b), as expected for a diffusion kinetic regime

225

Crystal Engineering of Metallic Nanoparticles

or for progressive nucleation,

a -2

jc (µA cm )

100

jc = zFci Di1/2 / 1/2 tc1/2 

Ed= 0.355 V Ed= 0.385 V

80



× 1 − exp−Ke Di No Ac td2 /2

jM 40 20 0

tM 0

50

100

150

200

td (s)

b 150 125 -2

jc (µA cm )

(3b)

60

100 75

jL tc

50 25 0 0.0

0.1

0.2

0.3 -1/ 2

0.4

0.5

0.6

-1/ 2

td (s ) Figure 5. Potentiostatic current transients (jc versus td  for palladium electrodeposition on HOPG using the same solution as in Figure 4. (a) Ed = 0355 V (o), Ed = 0.385 V (•). The current maximum jM at tM is indicated. (b) jc versus td−1/2 plot from current transient run at Ed = −0100 V. The straight line corresponds to Cottrell–Sand equation for linear diffusion. The arrows indicate the jL and tc values. Reprinted with permission from [41], Y. Gimeno et al., J. Phys. Chem. B 106, 4232 (2002). © 2002, American Chemical Society.

from the solution. As Ed is negatively increased jM increases and tM decreases. In all cases, after an Ed -dependent time tc that is usually less than 30 sec, free convection sets in and all transients tend to approach the quasisteady convectivediffusion regime (Figs. 4 and 5b). Therefore, depending on td , the kinetics of palladium electrodeposition changes from a linear diffusion to convective diffusion control.

In Equations (3a) and (3b), Di , and ci are the diffusion coefficient and the concentration of the electroactive species (i) in the bulk of the solution, respectively; No is the number of sites available for nucleation on the HOPG substrate; Ke = 8ci M/ 1/2 and Ke′ = 4/38ci M/ 1/2 ; M and are the molar mass and the density of the deposited metal, respectively; and Ac is the site-to-nucleous conversion frequency. Equations (3a) and (3b) reproduce the current transients for different metal electrodepositions. Figure 6 shows current transients for gold electrodeposition on HOPG and the comparison of these data to predictions from the models. More refined models including “clear field” effects around growing nuclei have been proposed [46, 47]. The nucleation rate (Ac No ) for progressive nucleation can be determined from the current density transient maxima resulting from relatively low cathodic overvoltages, c = Ed − Er . However, as very often current density maxima at high values of c are not observable at the time scale of conventional transient measurements, values of Ac No can be estimated from the Nc versus td plots derived from STM and AFM imaging, and SEM micrograph data, using equation [44, 45, 48] Nc = No 1 − exp−Ac td 

(4)

Nc being the number of nuclei at td . Data from palladium electrodeposition were plotted according to Equation (4) (Fig. 7). In this case, for −010 V < Ed < 0365 V, values of Ac No fulfill equation [19, 47] (Fig. 8) Ac No = K1 exp−K2 /c2 

(5)

In Equation (5), K1 depends on both the nucleous dimension (2-D, 3-D) and nucleation mechanism, as in, whether the metal lattice grows under diffusion or charge transfer 1.2

4. CRYSTAL GROWTH KINETICS

0.8

(3a)

0.6

2

Current transients similar to those described above for palladium electrodeposition have also been obtained for the electrodeposition of silver, palladium, gold, platinum, cadmium, and copper on HOPG from diluted aqueous solutions for td ≪ tc [13, 33, 41]. These transients often fit equations predicted by either instantaneous or progressive nucleation and 3-D growth models under diffusion control [42–45]. The jc versus td relationship can be expressed by the equations proposed by Scharifker and Hills [44], either for instantaneous nucleation,

j / jM

2

4.1. Initial Stages of Growth

jc = zFci Di1/2 / 1/2 td1/2 1 − exp−Ke Di No td 

inst prog Ed = 0.50 V

1.0

0.4 0.2 0.0 0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

t / tM

Figure 6. Comparison of normalized current transient data (jc /jM versus td /tM ) for gold electrodeposition on HOPG at Ed = 0500 V from aqueous 5 × 10−4 M gold chloride + 0.5 M sodium perchlorate + 1 × 10−2 M perchloric acid. Predictions from equations [3a and 3b] for the instantaneous (—-) and progressive (   ) nucleation and growth model under diffusion control are indicated.

226

Crystal Engineering of Metallic Nanoparticles

30

-2

-8

Nc (cm x 10 )

45

15

0

0

15

30

45

td (s) Figure 7. Nc versus td plot for palladium electrodeposition on HOPG from the same solution as in Figure 4 and Ed = −0100 V. The solid line corresponds to Eq. (4). Reprinted with permission from [41], Y. Gimeno et al., J. Phys. Chem. B 106, 4232 (2002). © 2002, American Chemical Society.

control. The constant K2 is related to Ac by the following relationship [48] K2 /c2 = Ac /kT

(6)

k being Boltzmann constant. Equation (6) implies a linear Ac No versus c−2 relationship (Fig. 8), and the value of K2 can be determined from the slope of this linear plot. Likewise, the number of atoms (n∗ ) involved in the formation of the critical nucleous can be calculated from the expression n∗ = 2Ac /zeo Ed − Er 

(7)

with eo = 160 × 10−19 Coulomb. Then, by combining Equations (6) and (7), it is possible to estimate values of n∗ at different Ed . For palladium electrodeposition it results in n∗ = 2 for Ed = 038 V, and n∗ = 0 for Ed = −01 V. Comparable values of n∗ have been reported for the electrocrystallization of several metals on foreign substrates [19]. In all these cases, the decrease in n∗ with Ed has been related to a decrease in the activation energy for nucleation with c . Thus, combined electrochemical and microscopy data can be used to obtain reliable information about the very early stages of growth of metal nucleous on HOPG. 30 25

ln A0 N0

20 15 10 5 0

0

5

10

15

20 -2

ηc

25

30

35

40

45

-2

(V )

Figure 8. Ln(Ac N0 ) versus 1/c2 plot to estimate the number of atoms at the critical nucleus (n*). Data obtained from the same solution as in Figure 4. Reprinted with permission from [41], Y. Gimeno et al., J. Phys. Chem. B 106, 4232 (2002). © 2002, American Chemical Society.

The preceding analysis points out that model predictions exclusively derived from electrochemical data should be checked against direct data from either STM or AFM imaging or SEM micrographs. In fact, for a number of metal electrodeposits, large discrepancies between values of No from imaging data and the number density of nuclei resulting from electrochemical data have been observed [32]. Likewise, for a particular system, the value of No depends on the carbon substrate. For silver electrodeposition on HOPG, the value of No increases from 57 × 106 to 3 × 107 cm−2 as Ed is shifted negatively, whereas values of No one order of magnitude smaller have been determined on mechanically polished vitrous carbon. For an isolated nucleous, Equations (3a) and (3b) imply a radial growth equation of the form r ≈ td1/2 , where r is the average radius of the growing nucleous. This equation is obeyed by data from the initial stages of platinum electrodeposition on HOPG [38]. Furthermore, for td > tM the current decay obeys a jc ≈ td−1/2 relationship (Fig. 6) as expected for pure diffusion. This behavior for jc , which is observed in the range tM ≤ td ≤ tc , has also been found for the electrodeposition of different metals from diluted solutions [32, 33, 41, 42]. Deviations from the jc ≈ td1/2 behavior are observed for td ≥ tc , as in when jc tends asymptotically to jL .

4.2. Electrodeposition Parameters: Shape and Size of Metal Crystals The initial stages of metal electrodeposition on HOPG have been extensively studied by in-situ and ex situ STM and AFM combined with electrochemical techniques [19, 30, 31]. When electrodeposition proceeds close to equilibrium conditions, such as when Ed is slightly more negative than Er , and a small charge density is involved, in the initial stages already metallic agglomerates (clusters) 1–2 nm in average diameter preferentially nucleated at step edges are produced. Under these conditions, there is agreement that the electrodeposition process takes place by the Volmer–Weber [20] growth mechanism (Fig. 1). For silver electrodeposition from acid plating solutions, the appearance of small ordered metallic domains has also been observed near step edges, probably acting as precursors to clustering [30]. For copper electrodeposition on HOPG, the formation of periodic arrays of clusters 2 nm in diameter following step edges has been observed [24]. Similar results have been obtained for platinum electrodeposition from acid plating baths on the same substrate [31], yielding clusters of average diameter d that roughly increases with td as d ≈ td1/2 [49]. The aspect ratio (f ) defined as the average height-to-average radius ratio is always small, f = h / r ≈ 01. The increase in either td or qd as Ed is made slightly more negative than Er , as in still maintaining a slow growth rate, results in the development of larger crystals that tend to develop regular patterns. Thus, at this stage, silver electrodeposition on HOPG yields triangular islands with closely packed (111) faces [31], whereas palladium electrodeposition on the same substrate produces quasi-hexagonal crystals [41]. Under comparably applied potential conditions [50], tetragonal tin crystals on both vitreous carbon and on

Crystal Engineering of Metallic Nanoparticles

227

HOPG were grown from aqueous tin sulphate-sulphuric acid solution. The increase in metal electrodeposition rate dramatically affects the shape and density of growing crystals. As a typical example, let us consider palladium electrodeposits on HOPG obtained for qd = 5 mC/cm2 as the electrodeposition rate is increased by shifting Ed negatively (Fig. 9). Thus, for Ed = 0380 V (Ed < Er ), faceted hexagonal crystals with r = 200–500 nm are grown (Fig. 9a) [41]. The corresponding crystal density evaluated from SEM micrographs is Nc = 04 × 108 cm−2 . For Ed = 0125 V, Nc increases to 36 × 108 cm−2 and r decreases markedly (Fig. 9b). The size distribution of these crystals is rather homogeneous with r  80 nm. The average height of these crystals measured from STM images is h = 35 nm. Accordingly, the value f is still low enough to describe crystals as compact rounded disks rather than hemispherical crystals. On the other hand, for Ed = −01 V (Ed ≪ Er ), such as when Ed is set in the potential range just preceding the hydrogen evolution reaction threshold potential, a large number, Nc = 4 × 109 cm−2 , of small branched palladium crystals with r = 70 nm is formed (Fig. 9c). The detailed structure of these small islands can only be resolved by STM imaging (Fig. 9d). A similar dependence of the crystal shape and size on Ed has also been observed for gold electrodeposits on HOPG [33]. The growth of the branched gold and palladium crystals for Ed = −01 V and different values of qd was followed by sequential STM imaging. It should be noted that qd ∝ td . Thus, for qd = 1 mC/cm2 (Fig. 10a), irregular crystals with protrusions either randomly distributed over HOPG terraces or aligned and coming into contact with neighboring crystals, decorating HOPG steps are observed. The palladium crystals grown along step edges form metallic mesowires that have been recently used as hydrogen sensors [25]. Higher

Figure 9. SEM micrographs of palladium electrodeposits on HOPG obtained from the same solution as in Figure 4; qd = 5 mC/cm2 . (a) Ed = 0380 V. (b) Ed = 0125 V. (c) Ed = −0100 V. (d) STM image (280 × 280 nm2 ) of palladium branched islands (Ed = −0100 V) qd = 3 mC/cm2 . The bar corresponds to z = 40 nm. Reprinted with permission from [41], Y. Gimeno et al., J. Phys. Chem. B 106, 4232 (2002). © 2002, American Chemical Society.

Figure 10. STM images of palladium islands on HOPG electrodeposited at Ed = −0100 V from the same solution as in Figure 4. (a) 15 × 15 m2 ; qd = 1 mC/cm2 . (b) Higher resolution 3-D STM image (156 × 156 nm2 , qd = 1 mC/cm2 , z = 40 nm, and cross section. (c) 023 × 023 nm2 STM image, qd = 15 mC/cm2 , z = 40 nm, and cross section. (d) STM image of branched islands (125 × 125 m2 , qd = 3 mC/cm2 , z = 40 nm. (e) 3-D STM image of a triangular island (025 × 025 m2 ), qd = 3 mC/cm2 , z = 40 nm. (f) 3-D detail and cross section of a palladium main branch, qd = 3 mC/cm2 , z = 40 nm. (g) Branched islands (2 × 2 m2 , qd = 4 mC/cm2 , z = 200 nm. Reprinted with permission from [41], Y. Gimeno et al., J. Phys. Chem. B 106, 4232 (2002). © 2002, American Chemical Society.

228

4.3. Quantitative Kinetic Data from Different Stages of Growth Quantitative kinetic data for the different stages of metal crystal growth can be obtained from the time dependence r and h derived by sequential STM and AFM images, as well as SEM micrographs. For the latter, in the case of 2-D-type deposits, only values of r can be derived. In general, both r and h fulfill power law dependences with

100 -2

1mC cm RSDdia = 6.4 = 73.5

a % Count

80 60 40 20 0

0

20 40 60 80 100 120 140 160 180 200

Particle diameter (nm) 100 -2

% Count

2 mC cm RSDdia = 11.9 = 148.9

b

80 60 40 20 0

0

20 40 60 80 100 120 140 160 180 200

Particle diameter (nm) 100 -2

c

3 mC cm RSDdia = 16.2 = 155.8

80

% Count

resolution STM images (Fig. 10b) show that each metallic island consists of an agglomeration of small quasispherical particles with 5 ≤ d ≤ 15 nm forming a 3-D core. For qd = 15 mC/cm2 instabilities at crystal edges can be seen (Fig. 10c), and for qd = 3 mC/cm2 , a net distinction between the central 3-D core and the 2-D branched pattern originated by instabilities around the core is observed (Fig. 10d). The quasi-2-D branched pattern is revealed by the fact that the maximum height of crystals remains almost the same as that of the 3-D core initially formed. Consistent with a (111) metallic lattice, the 2-D pattern around the core shows a net tendency to develop triangular crystals with most of the 2-D branches merging at 120 (Fig. 10e). Similar gold and palladium islands with preferential orientation (111) were grown on HOPG by vapor phase deposition at room temperature [51]. As recently reported [52, 53], STM images of palladium triangular crystals on HOPG produced from the vapor phase deposition demonstrate that for every stage before and after heating, irrespective of the degree of surface coverage, the deposit grows epitaxially in the fcc(111) direction. For palladium, secondary branches are separated by thin “fjords” 10 nm in width emerging from each main branch that have a tendency to produce a dense growth pattern. These patterns closely resemble those dense radial morphologies that have been observed for several metallic electrodeposits with a non-fractal mass and a self-affine fractal perimeter [54, 55]. Considering the crystal as a 2-D object, the fractal dimension (DF ) of branched crystal edge is DF = 13. On the other hand, gold crystals tend to produce dendritic patterns with DF = 176 [55]. Procedures for calculation of the fractal dimension of metallic crystals are given in [54–58]. For both gold and palladium electrodeposits, the cross section of crystals taken from the core border down to the tip of a main branch (Fig. 10f) approaches a 2-D profile with terraces almost atomically smooth (Fig. 10f, cross section). The height of each main branch decreases only about 5 nm in going from the core border to the branch tip with an average slope close to 3–4 . These tapering effects of the main branches mostly involve steps one atom high. Typical branch tip radii are close to 10 nm, whereas the tip height fluctuates between 2 and 4 nm. Finally, for qd = 4 mC/cm2 (Fig. 10g), the increase in both Nc and r leads to the partial overlap of neighbor branch centers. The full width at one-half maximum height value (FWHM) from the crystal size distribution function of these palladium crystals increases with qd (Fig. 11a–c) [13–14]. Thus, the relative standard deviation (RSD) in crystal diameter increases from 8.7% to 15.4% as d increases from 70 nm (qd = 1 mC/cm2 ) to 140 nm (qd = 3 mC/cm2 ).

Crystal Engineering of Metallic Nanoparticles

60 40 20 0

0

20 40 60 80 100 120 140 160 180 200

Particle diameter (nm) Figure 11. Particle size histograms for palladium electrodeposits. Data taken from Figure 10. (a) qd = 1 mC/cm2 . (b) qd = 2 mC/cm2 . (c) qd = 3 mC/cm2 .

time, namely, r ∝ tdn and h ∝ tdm , respectively. The values of the coarsening exponent, n and m, depend on Ed and qd as the relative contribution of different rate processes to crystal growth changes with Ed and td . Correspondingly, for m = n the value of f turns into a function of td , that is, f td  = h td / r td . Values of n and m for gold and palladium are assembled in Table 1.

Table 1. Values of m, n, and f resulting from the electrochemical growth of palladium, gold, silver and platinum particles on HOPG at 298 K. Palladium

Gold

Platinum

Silver

m

Ed < Epzc : 0.15

Ed > Epzc : 0.40 Ed < Epzc : 0.25





n

Ed < Epzc : 0.40 Ed > Epzc : 0.40

Ed > Epzc : 0.40 Ed < Epzc : 0.25

1/2

1/3

f

f t

0.3

0.1



Reference

[41]

[33]

[49]

[13–14]

Crystal Engineering of Metallic Nanoparticles

For the early stages of a hemispherical or spherical cluster grown under diffusion from the solution side and td ≤ tc , a r ∝ t n , dependence on n = 1/2 should be expected [44]. In this case f = 1, irrespective of td . On the other hand, for a 2-D-shaped cluster the same radial growth kinetic equation is obeyed, although in this case f td  approaches zero as td increases. At later stages, for td ≥ tc , as in when crystal growth is under a convective-diffusion regime, growth models predict a r ∝ t n relationship with n = 1/3 [2]. Conversely, when particle growth proceeds under surface reaction kinetics, a r ∝ t relationship is fulfilled [56–58]. Values of n for electrodeposition of various metals have been reported (Table 1). These figures for Ed ≈ Er are consistent with diffusion and convective-diffusion regimes [13, 14, 49], whereas for Ed < Er values of n and m ranging from 1/4 to 2/5 are found [33, 41]. Otherwise, for gold crystal growth f ≈ 03, a figure that is independent of td as n = m. The reverse situation occurs for the growth of palladium crystals as n > m, and f decreases rapidly with td . Values of f smaller than 1 indicate that a relatively fast surface metal atom relaxation takes place during crystal growth. It is well established that metal atom relaxation becomes a significant kinetic contribution for T ≪ 05Tm , Tm being the melting temperature of bulk metal. Therefore, it is reasonable to consider that under this situation surface diffusion of metal adatoms contributes to crystal growth [59]. This conclusion is supported by the value n = 2/5 [60, 61] that was determined for disk-shaped gold and palladium crystals grown at Ed more positive than Epzc , the potential of zero charge of the electrodepositing metal. For the growth of branched gold crystals at Ed < Epzc , the value n = 1/4 [62] also sustains a surface diffusion contribution to the process, although in this case there is a significant influence of activation energy barriers for surface diffusion via step edges on the global process [63], as discussed later.

4.4. Models for Crystal Growth In principle, as a first approach it seems reasonable to analyze the physical processes that determine the structure, shape, and size of crystals within the framework of the surface diffusion, deposition and aggregation (DDA) model [56–58]. This model is widely used for interpreting cluster growth data resulting from vapor phase deposition (Fig. 12a) [11, 64]. According to the DDA model, metal atoms impinging on the solid substrate diffuse as adatoms toward growing crystals to be captured at edges. Otherwise, adatoms impinging at the top of crystals diffuse downward to be captured at crystal edges. These processes occur without restrictions, as no activation energy barriers for adatom surface diffusion at step edges exist [63]. When a bump at a crystal edge is formed, it has a greater chance of capturing arriving atoms from the substrate rather than a reentrant, thus favoring tip growth and crystal branching (Fig. 12b) [56]. In the DDA model, tip splitting rather than tip stabilization becomes dominant, although tip stabilization resulting in dendritic patterns could also be produced by including the noise reduction technique in the model [65]. In principle, the DDA model could describe the growth of metal crystals on HOPG. For Ed → Er , the low flux of arriving particles implies a small nucleation rate (Equation (6))

229

Figure 12. Scheme of particle displacement in the DDA growth model. (a) Initial stages. (b) Advanced stages showing branching development.

and internuclei distances relatively large, so that the direct discharge of reactants on the susbtrate becomes possible. However, experimental compact disk-shaped island patterns with unstable edges that are formed for Ed → Er do not evolve according to the predictions of a DDA-like model including dendrite formation. On the other hand, when the surface diffusion of adatoms at crystal edges is enhanced by increasing the substrate temperature, it results in a transition from DDA-like to rounded compact, kinked islands [56–58]. These results, therefore, support the idea that for Ed → Er , metal adatom surface diffusion at crystal edges plays a key role in eliminating either the DDA or dendritelike structure. Another drawback in describing experimental results in terms of the DDA model lies in the fact that it predicts perfect 2-D patterns, as there is no restriction for moving adatoms on growing crystals to reach crystal edges. For gold electrodeposition, the value f = 03, irrespective of Er , indicates that gold adatoms produced by discharging AuCl− 4 ions on crystal tops must surmount an activation energy barrier at gold crystal steps [63]. A scheme showing the location and the relative magnitude of activation energy barriers for intra- and interterrace diffusion at metal surfaces is shown in Figure 13. The activation energy barrier at step edges is due to the decrease in the coordination number of adatoms along their interterrace trajectories across step edges. For barrierless surface diffusion, the development of a completely flat interface similar to that predicted by the Edwards–Wilkinson model [66] should be expected (n = 1/2). For gold adatom interterrace diffusion, the corresponding energy barrier is larger than that for intraterrace diffusion [67, 68]. Therefore, in this case, interterrace adatom diffusion becomes rate determining. Accordingly, a clear deviation in the crystal shape from a perfect 2-D pattern should be expected. It should be noted that for isotropic

230

Crystal Engineering of Metallic Nanoparticles

Figure 13. Qualitative scheme of Schwoebel–Erlich barrier for interterrace and intraterrace adatom surface diffusion. %Ek corresponds to the energy contribution to kink formation.

surface diffusion that implies the existence of a single activation energy barrier, the crystal evolves maintaining its original rounded or hexagonal shape. The height of the activation energy barrier at step edges is closely related to the value of n. Values n ≈ 2/5 have been related to low activation energy barriers at step edges [60, 61]. The influence of the latter in determining the departure of the crystal shape from a perfect 2-D pattern can also be derived from the analysis of f t values. Thus, as f t decreases with td , crystals tend to approach 2-D patterns as in the case of palladium (111) (Table 1). In this case, mean-field calculations have shown that the height of stepedge activation energy barriers is considerably lower than that for gold (111), so that for palladium (111) intraterrace diffusion becomes rate controlling [68]. From the above discussion, for Ed → Er , it can be concluded that any DDA-like model applicable to the growth of metal electrodeposits on HOPG should include the presence of activation energy barriers at step edges to account for shape evolution. Therefore, to explain the nearly compact disks found in this potential region, either enhanced edge surface diffusion and/or an isotropic activation energy barrier at steps should be included. A scheme of this type of crystal growth under isotropic adatom surface diffusion is depicted in Figure 14a. For Ed , which is more positive than the potential of zero charge of the depositing metal (Epzc ), the role played by metal adatom surface diffusion is sustained by the fact that the presence of a small amount of additive such as citric acid (c = 01 mM) in the plating solution eliminates branching and promotes the formation of rounded 3-D islands. As seen by in-situ STM imaging [69], this effect is caused by the adsorption of additive molecules at step edges, which hinders surface diffusion of gold atoms. On the other hand, for fcc(111) surfaces, two different step edges, A- and B-type, should be distinguished [70]. The height of the activation energy barrier for interterrace adatom surface diffusion across A-type step edges is greater than across B-type step edges. This anisotropy in interterrace adatom surface diffusion results in the growth of triangular crystals. A scheme for anisotropic surface diffusion is shown in Figure 14b. Experimental data from gold and palladium electrodeposition show that anisotropic surface diffusion is approached for Ed < Epzc . In this case, for Ed ≪

Figure 14. (a) Scheme of isotropic growth model. (b) Ball scheme of adatom surface diffusion across steps A and B on a (111) substrate. (c) Ball scheme of adatom surface diffusion at corner (1) and edges (2).

Er , as a large number of nuclei is initially formed (Equations (4)–(6)), most of the ions coming from the solution are discharged on crystal tops, and the descending flux of adatoms across B-type steps is favored. Therefore, triangular crystals are by far preferentially formed. Hence, the existence of activation energy barriers of different height at A- and B-step edges leading to anisotropic surface diffusion of metal adatoms can explain the island shape obtained for Ed < Epzc . For Ed ≪ Er , dendritic patterns that are observed for both palladium and gold crystals also indicate the presence of anisotropic activation energy barriers at step edges [70]. From an atomistic standpoint, it is possible to distinguish two surface diffusion processes at island edges, depending on whether an adatom starts from a site that is laterally twofold (edge diffusion) (Fig. 14c) or onefold coordinated to the crystal edge (corner diffusion) [70]. Hexagonal lattice dendritic patterns with trigonal symmetry have been generated introducing anisotropic corner surface diffusion, whereas edge surface diffusion leads mainly to a thickening of island branching. For these surface diffusion processes, the symmetry of the substrate also plays a key role in the development of dendritic patterns [70]. This fact can explain, for instance, why under comparable experimental conditions, no dendritic crystals for gold electrodeposition onto amorphous carbon can be observed [71]. The preceding analysis provides a reasonable explanation to results obtained for Ed < Epzc , but the question of why anisotropic surface diffusion is no longer observed for Ed > Epzc still remains open. In this case, an answer can be given after paying attention to the influence of the applied electric potential on surface diffusion processes.

4.5. The Electric Potential Influence For metal electrodeposition processes, the height of activation energy barriers for adatom surface diffusion across step edges depends on the applied electric potential. Accordingly, it is possible to control the nanostructure and shape of metal

231

Crystal Engineering of Metallic Nanoparticles

islands by properly adjusting the electrodeposition conditions, particularly the applied electric potential in potentiostatic experiments. For a potential-dependent metal island shape on HOPG, the adsorption of anions plays an important role. It is well known that anions in solution interact with the metal surface either electrostatically or specifically adsorbed. Both interactions depend on the electric potential applied to the electrochemical interface [72]. In principle, for Ed < Epzc , the degree of metal surface coverage by adsorbed anions is negligible, as in this case the metal surface is negatively charged with respect to the solution. The reverse situation occurs for Ed > Epzc . Then, as Ed is shifted positively, the electron density at the metal surface decreases and the degree of surface coverage by anions increases, eventually leading to the formation of a welldefined ordered anion adlattice [73]. For example, for either chloride or metal-chloride complex-containing plating solution, at Ed > Epzc , anions accumulate at the metal/solution interface in the course of metal electrodeposition, and either chloride or metal-chloride complex ions become specifically adsorbed at growing metal islands. The global metal electrodeposition reaction from chloride containing metal anions is represented by Equation (1b) for Me = palladium, X = chlorine, and 1 ≤ x ≤ 6. When Ed < Epzc , as the metal surface is negatively charged with respect to the solution, chloride ions produced by Equation (1b), as well as anions in general, diffuse outward from the metal/solution interface, and therefore, the influence of anion adsorption on metal electrodeposition becomes negligible. Then, the metal deposit approaches a dendritic growth pattern, as expected when an anisotropic corner diffusion regime operates. It is interesting to note that the same growth pattern has been observed for vapor phase deposition of silver islands on platinum (111) [70]. Conversely, at Ed > Epzc , the metal being positively charged, chloride anions produced by Equation (1b) become specifically adsorbed on the growing crystal surface. Under these particular conditions, it is well-established that specific adsorption phenomena are enhanced at step edges. The increase in chloride ion adsorption with Ed correlates with an increase in the surface diffusion coefficient of metal adatoms, leading to an enhancement of both adatom surface mobility [72] and step displacement [73]. In fact, when Ed is increased from 0.30 V (Epzc ≈ 025 V versus SCE) to 0.80 V, the surface diffusion coefficient of gold adatoms on gold immersed in acid solution increases from Ds = 1 × 10−12 cm2 /s to Ds = 8 × 10−12 cm2 /s [74]. Hence, the specific adsorption of anions at step edges produces a decrease in the height of energy barriers for adatom surface diffusion, that is, isotropic surface diffusion conditions are then approached that favor the growth of kinked rounded crystals. For gold island electrodeposition, this explanation is consistent with the change of n from 1/4 to 2/5 [75]. On the other hand, values of Ds for palladium adatom surface diffusion on palladium, either in the presence or in the absence of chloride ions are not available yet, but they can be estimated from the Gjostein–Rhead equation [76–78] Ds = 749 × 10−4 exp−30Tm /RT  + 0014 × 10−4 exp−13Tm /RT 

(8)

In Equation (8), Tm is the melting point of the surface species, and R = 1987 cal/mol K. Let us consider Equation (8) to determine whether the most likely surface-diffusing species is closely related to palladium adatoms or to palladium-chloride-complex adsorbates. Based on electrochemical kinetic data [41], let us assume that for Ed > Epzc , a PdCl2 adsorbate on palladium is formed [77, 78]. Then, from the application of Equation (8) for palladium (Tm = 1827 K) and PdCl2 (Tm ≈ 773 K), it results in Ds = 13 × 10−14 cm2 /s for PdCl2 , and Ds = 57 × 10−24 cm2 /s for palladium at 298 K. Therefore, we can conclude that a palladium-chloride complex species [79] is responsible for the surface relaxation process occurring during the palladium crystal electrochemical growth. The increase in the value of Ds for palladium-chloridecomplex-ion-containing solution follows the same tendency that has already been reported for the surface mobility of gold adatoms on gold under comparable conditions [74]. For such a high value of Ds , the average traveling rate of [PdCl2 ] species on the palladium surface is about 1 nm/s. This figure is consistent with the surface relaxation rate that contributes to the formation of rounded islands for Ed > Epzc . It should be noted that the high value of Ds referred to above is of the same order of magnitude as that of Ds that has recently been found for the surface diffusion of disk-shaped aggregates of cadmium sulfide nanocrystals on HOPG at 50–100  C [80]. The role of chloride ions in the surface diffusion rate was confirmed by increasing their concentration in the plating solution. In fact, the addition of chloride salt promotes 2-D growth, that is, it decreases the value of f , and tends to hinder dendritic growth [69] as found from AFM imaging for the growth of flat cobalt crystals on HOPG by electrodeposition from chloride-containing solutions [50].

5. APPLICATIONS 5.1. Nanowires Among the different methods for fabricating long metallic nano- and mesowires, the most successful ones are template synthesis [81–83] and step-edge decoration [84–87]. The electrochemical method for preparation of metallic meso- and nanowires is based on the selective electrodepositon of the metal directly at step edges utilizing HOPG substrate. This step-edge decoration method has been proposed by Liu et al. [13], Penner [14], Penner et al. [25], and Zach et al. [26] from either metal or metal oxide electrodeposition on HOPG. For the preparation of palladium meso- and nanowires, the solid solution interface is perturbed by a routine consisting of three successive potential steps. The first step is adjusted to oxidize step edges on HOPG just prior to deposition; the second step, which involves a large cathodic overvoltage for metal deposition of about 0.3 V for 5–10 ms, favors metal nucleation; the third step involves a low cathodic overvoltage for metal growth of less than 0.1 V. The radius of the nanowires is controlled by the electrodeposition time [88, 89]. Metallic palladium meso- and nanowires can also be produced by a two-stage electrochemical process involving first the potentiostatic electrodeposition of palladium followed

232 by a controlled anodic stripping stage. A typical palladium nano/mesowire array prepared by this method is shown in Figure 15. The preparation of metallic nanowires also proceeds via the electrodeposition of a metal oxide. This is the case of the electrodeposition of molybdenum dioxide on HOPG from dilute molybdate solution at pH 6.5. This process is followed by hydrogen reduction at 500  C yielding HOPG-supported metallic molybdenum meso- and nanowires [88, 89]. Nanoparticles and nanowires are potentially very efficient chemical sensors. The chemical sensing mechanism is envisioned as the adsorption and eventually diffusion of the analyte species at the surface inducing there a change in a measurable property. An example of a chemical sensor operating on the above principle, which has been described recently [88, 89], is that of gold nanowires showing changes in the electrical conductivity upon exposure to thiols of amines. Arrays of mesoscopic palladium wires have been applied as hydrogen sensors and hydrogen-actuated switches [25]. These devices have been constructed by transferring the palladium nanowire arrays from HOPG onto a cyanoacrylate or polyurethane film. In contrast to hydrogen sensors based on macroscopic palladium resistors, which become more resistive in the presence of hydrogen, the resistance of the palladium mesowire array decreases under hydrogen. From AFM and SEM observations, the likely mechanism that has been proposed for the operation of the palladium mesowire array to account for its response as a hydrogen-actuated switch involves the closing of nanoscopic break junctions along the step edges caused by the dilation of wire grains undergoing the reversible conversion of palladium to palladium hydride. Accordingly, in the absence of hydrogen, each break junction reverts to open circuit [25].

5.2. Pattern Transfer to Polymeric Materials HOPG patterned with electrodeposited nanometer-sized metallic crystals has been used as a template to fabricate nanocavities on polymeric materials [27]. The procedure for

Figure 15. SEM micrograph of palladium islands on HOPG electrodeposited at Ed = −0100 V, from the solution indicated in Figure 4, for qd = 5 mC/cm2 , after anodic stripping in 0.1 M aqueous perchloric acid to reach qd = 13 mC/cm2 .

Crystal Engineering of Metallic Nanoparticles

pattern transfer is simple. It consists of pouring a prepolymer solution onto the patterned HOPG surface. After evaporation, a polymer film that can be easily released from the HOPG surface is formed. The inner face of the polymer film, which has been in contact with the HOPG surface, exhibits nanocavities—the pattern transfer from the HOPG surface to the polymeric material takes place. It should be noted that this simple method to produce nanocavities onto polymeric surfaces has also been performed using either patterned silicon or patterned metallic substrates. However, in these cases, the template surface has to be modified by a self-assembled monolayer of alkanethiols (metals) or silanes (silicon) in order to release the polymer film from the template without damage. The advantage of using patterned HOPG arises from the fact that, due to its extremely low chemical reactivity, no surface chemistry modification is needed to avoid damage. Note that depending on the metalpolymer interactions, the electrodeposited metal particles can either be transferred to the polymeric film (as described in Section 5.1) or remain on the HOPG surface forming cavities on the polymer (nanocavity fabrication).

5.3. Heterogeneous Catalysis The reactivity of the nanometer-sized compact disks and branched palladium particles for hydrogen atom electroadsorption/electrodesorption reactions in aqueous 0.1 M perchloric acid was investigated by cyclic voltammetry at v = 01 V/s [90]. For the sake of comparison, blanks were run that utilized massive either palladium or platinum electrodes. For massive palladium, the voltammogram recorded from 0 V to −03 V (Fig. 16a) shows a cathodic current, essentially related to hydrogen absorption by palladium, and to some extent the initial formation of molecular hydrogen at higher cathodic overvoltage [91]. The reverse potential scan shows the electro-oxidation of absorbed hydrogen as the main anodic process with a broad anodic current peak at −0075 V. This voltammogram shows no trace of adsorbed hydrogen as might be expected at −030 V and −015 V (Fig. 16a, inset). Therefore, in contrast to the voltammogram of polycrystalline platinum that exhibits conjugated pairs of current peaks related to weakly and strongly bound hydrogen adatoms at approximately −005 V and −020 V, respectively, no hydrogen atom electroadsorption/ electrodesorption can be concluded from the voltammogram of smooth massive palladium. Comparable voltammograms were recorded for nanometer-sized palladium crystals of different morphologies on HOPG, namely, compact disk-shaped crystals prepared at Ed = 0125 V (qd = 5 mC/cm2  (Fig. 16b), and branched particles produced at Ed = −0100 V (qd = 5 mC/cm2 ) (Fig. 16c). In contrast to massive palladium, both types of palladium islands exhibit voltammetric features resembling those for hydrogen atom electroadsorption/ electrodesorption on platinum [92]. For palladium crystals supported on HOPG electrodes, these processes are characterized by peaks I and II at approximately −025 V and −015 V, respectively (Fig. 16c), and the anodic-to-cathodic voltammetric charge ratio approaches 1 as expected for relatively fast surface electrochemical processes [93, 94]. It can also be observed that for branched palladium islands

233

Crystal Engineering of Metallic Nanoparticles

HOPG, particularly on branched crystals. It should be noted that the voltammogram resulting from a palladiumplated HOPG electrode at Ed = −0100 V (qd = 20 mC/cm2 ) shows a contribution of hydrogen electroadsorption/electrodesorption much smaller than that observed for branched palladium crystals [95]. To confirm the high electrochemical reactivity of nanometer-sized branched palladium crystals, the electrochemical oxidation of hydrazine on massive palladium, nanometer-sized compact disk-shaped particles and nanometer-sized branched palladium crystals (qd = 5 mC/cm2 ) supported on HOPG were also investigated [90]. For this purpose, aqueous 6 × 10−3 M hydrazine + 01 M potassium sulphate at 298 K was employed. For massive palladium (Fig. 17a), the voltammogram shows a broad anodic current peak at 031 ± 001 V related to the electro-oxidation of hydrazinium ions [90]. This peak is followed by another broad peak at approximately 0.6 V, which corresponds to the formation of the oxygen-containing layer on palladium [96, 97]. The latter is electroreduced as the potential scan is reversed, as seen by the broad current peak at approximately 0.5 V. For nanometer-sized compact disk crystals produced at Ed = 0125 V (Fig. 17b), the anodic current peak potential related to hydrazinium ion electro-oxidation is 033 ± 001 V, whereas for branched palladium crystals (Fig. 17b) it is 026 ± 001 V. In this case, the conjugated cathodic reaction can also be observed at the same potential during the reverse potential scan. Furthermore, the initial portion of the voltammogram fits a linear Ed versus ln jc

0.8

a

0.0 15 0 0 -2

-0.4

j / ∝A cm

-2

j ( µA cm )

0.4

-0.8

-1 5 0 -3 0 0 -4 5 0 -0 .2

-0 .1

0 .0

0 .1

0 .2

0 .3

E /V

-1.2 -0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

E(V vs SCE) 400

b

II

-2

-1

j ( µA cm µg )

I 200 0 -200 -400 -600 -0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

E(V vs SCE)

c

200 0

4000 -200

3200

j (µΑ cm )

-400

-2

-2

-1

j ( µA cm µg )

400

-600 -0.3

-0.2

-0.1

0.0

0.1

0.2

a

2400 1600 800

0.3

E(V vs SCE)

0 -0.4 -0.2

0.0

0.2

0.4

0.6

0.8

1.0

E (V vs SCE) 4000

-2

-1

j (µΑ cm HOPG µg Pd)

Figure 16. Current density versus potential curves recorded at 0.1 V/s from 0.30 to −030 V to reach the region where hydrogen atom electrosorption/electrodesorption reactions are observed. Aqueous 0.1 M perchloric acid. (a) Polycrystalline palladium electrode (0.125 cm2 geometric area). The voltammogram (inset) corresponds to the cathodic potential limit −0.15 V. (b) Quasi-hemispherical palladium islands on HOPG, Ed = 0125 V, qd = 5 mC/cm2 . (c) The continuous trace corresponds to branched palladium islands on HOPG, Ed = −010 V, qd = 5 mC/cm2 . 298 K. Electrodeposits have been obtained as indicated in Figure 9. Reprinted with permission from [90], Y. Gimeno et al., Chem. Mater. 13, 1857 (2001). © 2001, American Chemical Society.

E d= 0. 125V E d= -0.1 V

b

3200 2400 1600 800 0 -0.4 -0.2

0.0

0.2

0.4

0.6

0.8

1.0

E (V vs SCE)

supported on HOPG, the peak I/peak II charge ratio is increased, as compared to that resulting from compact diskshaped crystals. Therefore, the contribution of the hydrogen adatom electrodeposition reaction PdHad = Pd + H+ + e−

(9)

as compared to hydrogen absorption is considerably enhanced on nanometer-sized palladium crystals supported on

Figure 17. Current density versus potential curves for hydrazine electro-oxidation recorded at 0.2 V/s. Aqueous 6 × 10−3 M hydrazinium sulphate + 0.1 M potassium sulphate. (a) Polycrystalline palladium electrode. (b) Quasi-hemispherical, Ed = 0125 V (full trace), and branched, Ed = −010 V (open circles) palladium islands on HOPG (qd = 5 mC/cm2 ). 298 K. The current density in (b) is also referred to the weight of electrodeposited palladium calculated from qd . Reprinted with permission from [90], Y. Gimeno et al., Chem. Mater. 13, 1857 (2001). © 2001, American Chemical Society.

234

Crystal Engineering of Metallic Nanoparticles

plot, with the slope %Ed /% ln jc decreasing from 015 ± 001 to 009 ± 001 V/ln unit in going from smooth to branched crystals. These results are consistent with electrochemical kinetics under intermediate control [98]. Under the above-mentioned conditions, the voltammetric electrochemical oxidation of hydrazine is a 4-electron reaction N2 H5+

+

= N2 + 5H + 4e



(10)

with a pair of conjugated peaks at the range 0.26–0.33 V. This complex electrochemical reaction depends on the crystallography, state, and history of the electrode surface, as has been concluded for platinum electrodes in the same solution [99]. The decrease in the value of the slope %Ed /% ln jc , the increase in height of conjugated peaks, and the displacement of their peak potentials by 0.07 V negatively indicate an increased catalytic effect of nanometer-sized branchedpalladium crystals on HOPG-supported electrodes. Therefore, electrodes behave as the electrochemically most active material for both test reactions.

6. CONCLUSIONS The electrodeposition of metallic crystals on carbon HOPG has been reviewed, in particular, in relation to crystal structure and shape. At the mesoscale, narrow particle size distributions of metal crystals can be obtained by conventional potentiostatic electrodeposition of metal ions from aqueous solutions. The shape and structure of mesocrystals depend on the applied electric potential. Over a wide potential range, the aspect ratio of crystals differs considerably from that of hemispherical particles. This departure can be explained by growth models, including surface adatom diffusion and step-dependent activation energy barriers at step edges. Anions adsorption from the solution plays a key role in determining the height of activation energy barriers for surface diffusion. The electrocatalytic activity of branched mesocrystals for different processes is considerably greater than that of compact disk-shaped mesocrystals of comparable size. Metallic electrodeposition on carbons can also be used for nano/mesowire fabrication with potential applications as chemical sensors, and promising interconnectors for nanoelectronic devices and polymer embossing. The discussion of these issues is important for facing problems related to the crystal engineering of metalic nanoparticles.

Voltammogram Current versus potential record obtained under a triangular potential scan between an upper and a lower switching potential.

ACKNOWLEDGMENTS This work was financially supported by projects PICT 995030 and PICT 98 N 06-03251 from Agencia Nacional de Promoción Científica y Tecnológica (Argentina), PIP 0897 from Consejo Nacional de Investigaciones Científicas y Técnicas (Argentina), IN94-0553 and PB94-0592A from DGICYT (Spain), and PI1999-128 from Gobierno de Canarias (Spain). R.C.S. thanks the Dirección General de Investigación Científica y Enseñanza Superior, SAB20000157 (Spain).

REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

GLOSSARY Exchange current density Value of the current density at the equilibrium potential of the electrochemical interface. Mesoparticles Particles with size comprised between 100 to 1000 nm. Nanoparticles Particles with size up to 100 nm. Overvoltage The potential difference between the applied electric potential and the reversible potential of the electrochemical interface. Potentiodynamic polarization curve A current versus potential record obtained under a single linear potential scan.

23.

24. 25. 26. 27. 28. 29.

G. Frens, Nature 20, 241 (1973). A. Henglein, Chem. Rev. 89, 1861 (1989). M. T. Reetz and W. J. Helbing, J. Am. Chem. Soc. 116, 7401 (1994). T. S. Ahmadi, Z. L. Wang, T. C. Green, A. Henglein, and M. A. El Sayed, Science 272, 1924 (1996). T. S. Ahmadi, Z. L. Wang, A. Henglein, and M. A. El-Sayed, Chem. Mater 8, 1161 (1996). A. Henglein and M. Giersig, J. Phys. Chem. B 103, 9533 (1999). M. B. Mohamed, Z. L. Wang, and M. A. El-Sayed, J. Phys. Chem. A 103, 10255 (1999). G.-Q. Lu and A. Wieckowski, Colloid and Interface Science 5, 95 (2000). V. F. Puntes, K. M. Krishnan, and A. P. Alivisatos, Science 291, 2115 (2001). A. Henglein, J. Phys. Chem. B 104, 1206 (2000). C. Binns, Surf. Sci. Rep. 44, 1 (2001), and references therein. C. R. Henry, Appl. Surf. Sci. 164, 252 (2000). H. Liu, F. Favier, K. Ng, M. P. Zach, and R. M. Penner, Electrochim. Acta 47, 671 (2001). R. M. Penner, J. Phys. Chem. B 106, 3339 (2002). K. Itaya, S. Sugawasa, and K. Higaki, J. Phys. Chem. 92, 6714 (1988). M. Aindow and J. P. G. Farr, Trans. Inst. Mater. Finish. 70, 171 (1992). A. R. Layson and M. R. Columbia, Microchem. J. 56, 103 (1997). P. Garrido, E. Gomez, and E. Valles, J. Electroanal. Chem. 441, 147 (1998). E. Budevski, G. Staikov, and W. J. Lorenz, “Electrochemical Phase Formation and Growth.” VCH, Weinheim, Germany, 1996. M. Volmer and A. Z. Weber, Phys. Chem. 119, 277 (1926). J. N. Stranski and L. Krastanov, Ber. Akad. Wiss. Wien 146, 797 (1938). F. C. Frank and J. H. van der Merwe, Proc. R. Soc. London A 198, 205 (1949). G. A. Somorjai, “Introduction to Surface Chemistry and Catalysis.” John Wiley & Sons, New York, 1994; N. Memmel, Surf. Sci. Rep. 32, 91 (1998). T. Tanii, K. Ishibashi, K. Ohta, and I. Ohdomari, Appl. Surf. Sci. 162–163, 662 (2000). R. M. Penner, F. Favier, E. C. Walter, and M. P. Zach, Science 293, 2227 (2001). M. P. Zach, K. H. Ng, and R. M. Penner, Science 290, 2120 (2000). P. L. Schilardi, O. Azzaroni, and R. C. Salvarezza, Nano-tech. submitted. “Scanning Tunneling Microscopy I,” edited by H.-J. Guntherodt and R. Wiesendanger, Springer-Verlag, Berlin, 1992. M. Paunovic and M. Schlesinger, in “Fundamentals of Electrochemical Deposition.” Wiley, New York, 1998.

Crystal Engineering of Metallic Nanoparticles 30. L. Vázquez, A. Hernández-Creus, P. Carro, P. Ocón, P. Herrasti, C. Palacio, J. M. Vara, R. C. Salvarezza, and A. J. Arvia, J. Phys. Chem. 96, 10454 (1992). 31. J. L. Zubimendi, L. Vázquez, P. Ocon, J. M. Vara, W. E. Triaca, R. C. Salvarezza, and A. J. Arvia, J. Phys. Chem. 97, 5095 (1993). 32. M. Miranda-Hernández, I. González, and N. Batina, J. Phys. Chem. 105, 4212 (2001). 33. H. Martin, P. Carro, A. Hernández Creus, S. González, R. C. Salvarezza, and A. J. Arvia, Langmuir 13, 100 (1997). 34. K. Shimazu, D. Weisshaar, and T. Kuwana, J. Electroanal. Chem. 223, 223 (1987). 35. K. Shimazu, K. Uosaki, H. Kita, and Y. Nodasaka, J. Electroanal. Chem. 256, 481 (1988). 36. J. V. Zoval, J. Phys. Chem. B 102, 1166 (1998). 37. A. C. Hill, R. E. Patterson, J. P. Sefton, and M. R. Columbia, Langmuir 15, 4005 (1999). 38. I. Lee, K.-Y. Chan, and D. L. Phillips, Appl. Surf. Sci. 136, 321 (1998). 39. S. Liu, Z. Tang, E. Wang, and S. Dong, Electrochem. Commun. 2, 800 (2000). 40. S. Chen and A. Kucernak, Electrochem. Commun. 4, 80 (2002). 41. Y. Gimeno, A. Hernández Creus, P. Carro, S. González, R. C. Salvarezza, and A. J. Arvia, J. Phys. Chem. B 106, 4232 (2002). 42. A. Hernández-Creus, A. E. Bolzán, P. Carro, S. González, R. C. Salvarezza, and A. J. Arvia, Electrochim. Acta 38, 1545 (1993). 43. A. Hernández-Creus, P. Carro, S. González, R. C. Salvarezza, and A. J. Arvia, Electrochim. Acta 37, 2215 (1992). 44. B. Scharifker and G. Hills, Electrochim. Acta 28, 879 (1983). 45. B. Scharifker and J. Mostany, J. Electroanal. Chem. 177, 13 (1984). 46. L. Heerman and A. Tarallo, J. Electroanal. Chem. 470, 70 (1999). 47. L. Heerman and A. Tarallo, Electrochem. Commun. 2, 85 (2000). 48. E. B. Budevski, in “Comprehensive Treatise of Electrochemistry” edited by B. E. Conway, J. O’M. Bockris, E. Yeager, S. U. M. Khan, and R. M. White, Vol. 7, Plenum Press, New York 1983, p. 399. 49. I. Lee, K.-Y. Chan, and D. L. Phillips, Ultramicroscopy 75, 69 (1998). 50. E. Gómez, E. Guaus, F. Sanz, and E. Vallés, J. Electroanal. Chem. 465, 63 (1999). 51. R. Q. Hwang, J. Schroder, G. Gunther, and R. J. Behm, Phys. Rev. Lett. 67, 3279 (1991). 52. I. Kojima, A. K. Srivastava, and M. J. Kurahashi, J. Appl. Phys. 30, 1852 (1991). 53. M. Bonet, E. Boschung, J. Hayoz, T. Pillo, G. Dieter, and P. Aebi, Surf. Sci. 473, 17 (2001). 54. “The Fractal Approach to Heterogeneous Chemistry,” edited by D. Avnir, John Wiley & Sons, New York, 1989. 55. J. M. Gómez-Rodríguez, A. M. Baró, L. Vázquez, R. C. Salvarezza, J. M. Vara, and A. J. Arvia, J. Phys. Chem. 96, 347 (1992). 56. A. L. Barabasi and H. E. Stanley, in “Fractal Concepts in Surface Growth.” Cambridge University Press, Cambridge, 1995, and references therein. 57. “Dynamics of Fractal Surfaces,” edited by F. Family and T. Vicsek, World Scientific, Singapore, 1991. 58. F. Family, Physica A 168, 561 (1990), and references therein. 59. H. P. Bonzel, in “Surface Chemistry of Materials,” edited by J. M. Blakeley, Academic Press, New York, 1975, p. 280. 60. P. Smilauer and D. D. Vvedenski, Phys. Rev. B 52, 14263 (1995). 61. S. Majaniemi, T. Ala-Nissila, and J. Krug, Phys. Rev. B 53, 8071 (1996). 62. M. Siegert and M. Plischke, Phys. Rev. Lett. 73, 1517 (1994). 63. R. Schwoebel, J. Appl. Phys. 40, 614 (1969). 64. J. Jacobsen, K. W. Jacobsen, S. Stoltze, and J. K. Norskov, Phys. Rev. Lett. 74, 2295 (1995). 65. P. Meakin, in “The Fractal Approach to Heterogeneous Chemistry,” edited by D. Avnir, John Wiley & Sons, New York, 1990, p. 131.

235 66. S. F. Edwards and D. R. Wilkinson, Proc. R. Soc. London A 381, 17 (1982). 67. G. Erlich, Surf. Sci. 299/300, 628 (1994), and references therein. 68. Y. Li and E. DePristo, Surf. Sci. 351, 189 (1996). 69. H. Martín, P. Carro, A. Hernández Creus, S. Gonzalez, G. Andreasen, R. C. Salvarezza, and A. J. Arvia, Langmuir 16, 2915 (2000). 70. H. Brune, H. Roder, K. Bromann, K. Kern, J. Jacobsen, P. Stoltze, K. Jacobsen, and J. Norskov, Surf. Sci. 349, L115 (1996), and references therein. 71. D.-I. Lu, Y. Okawa, M. Ichihara, A. Aramata, and K.-I. Tanaka, J. Electroanal. Chem. 406, 101 (1996). 72. D. J. Trevor, C. E. D. Chidsey, and D. N. Loiacono, Phys. Rev. Lett. 62, 929 (1989). 73. X. Gao, G. J. Edens, and M. J. Weaver, J. Electroanal. Chem. 376, 21 (1994), and references therein. 74. M. P. García, M. M. Gómez, R. C. Salvarezza, and A. J. Arvia, J. Electroanal. Chem. 347, 237 (1993). 75. H. Martin, P. Carro, A. Hernández Creus, S. González, R. C. Salvarezza, and A. J. Arvia, J. Phys. Chem. B 103, 3900 (1999). 76. N. A. Gjostein, in “Surfaces and Interfaces,” edited by J. J. Burke, N. L. Reed, and V. Weiss, Syracuse University Press, Syracuse, New York, 1967, p. 271. 77. G. E. Rhead, Surf. Sci. 15, 353 (1969). 78. G. E. Rhead, Surf. Sci. 22, 223 (1970). 79. L. G. Sillén, in “Stability Constant of Metal-Ion Complexes, Section 1: Inorganic Ligands,” The Chemical Society, Burlington House, London, 1964, p. 288. 80. G. Ge and L. E. Brus, Nano Lett. 1, 219 (2001). 81. C. A. Foss, M. J. Tierney, and C. R. Martin, J. Phys. Chem. 96, 9001 (1992). 82. C. K. Preston and M. Moskovits, J. Phys. Chem. B 97, 8495 (1993). 83. L. Sun, P. C. Searson, and C. L. Chien, Phys. Rev. B 61, R6463 (2000). 84. T. Jung, R. Schlittler, J. K. Gimzewski, and F. J. Himpsel, Appl. Phys. A 61, 467 (1995). 85. M. Blanc, K. Kuhnke, V. Marsico, and K. Kern, Surf. Sci. 414, L964 (1998). 86. R. J. Nichols, D. M. Kolb, and R. J. Behm, J. Electroanal. Chem. 313, 109 (1991). 87. S. Morin, A. Lachenwitzer, O. M. Magnussen, and R. J. Behm, Phys. Rev. Lett. 83, 5066 (1999). 88. F. Favier, E. C. Walter, M. P. Zach, T. Benter, and R. M. Penner, Science 293, 2227 (2001). 89. E. C. Walter, F. Favier, and R. M. Penner, Anal. Chem. 74, 1546 (2002). 90. Y. Gimeno, A. Hernández Creus, S. González, R. C. Salvarezza, and A. J. Arvia, Chem. Mater. 13, 1857 (2001). 91. R. C. Salvarezza, C. Montemayor, E. Fatas, and A. J. Arvia, J. Electroanal. Chem. 313, 291 (1991). 92. N. Hoshi, M. Kuroda, and Y. Hori, J. Electroanal. Chem. 521, 155 (2002). 93. S. Srinivasan and E. Giliadi, Electrochim. Acta 11, 321 (1966). 94. N. Baldauf and D. M. Kolb, Electrochim. Acta 38, 2145 (1993). 95. Y. Gimeno, A. Hernandez Creus, R. C. Salvarezza, A. J. Arvia, to be submitted. 96. A. Bolzan, M. E. Martins, and A. J. Arvia, J. Electroanal. Chem. 172, 221 (1984). 97. A. Bolzan, A. C. Chialvo, and A. J. Arvia, J. Electroanal. Chem. 179, 71 (1984). 98. A. J. Bard and L. Faukner, “Electrochemical Methods.” John Wiley & Sons, New York, 1960. 99. R. Gómez, J. M. Orts, A. Rodes, J. M. Feliú, and A. J. Aldaz, J. Electroanal. Chem. 358, 287 (1993).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Crystallography and Shape of Nanoparticles and Clusters J. M. Montejano-Carrizales Universidad Autónoma de San Luis Potosí, México

J. L. Rodríguez-López Instituto Potosino de Investigación Científica y Tecnológica, San Luis Potosí, México

C. Gutierrez-Wing Instituto Nacional de Investigaciones Nucleares, Mexico, and University of Texas at Austin, Austin, Texas, USA

M. Miki-Yoshida University of Texas at Austin, Austin, Texas, USA, and Centro de Investigación en Materiales Avanzados, Chichuahua, México

M. José-Yacaman University of Texas at Austin, Austin, Texas, USA, and Instituto de Física, UNAM, México

CONTENTS 1. Introduction 2. Geometric Considerations 3. Clusters and Nanoparticles with Pentagonal Symmetry 4. Synthesis of Nanoparticles 5. Transmission Electron Microscopy (TEM) and Simulations of Images of Nanoparticles 6. Conclusions Glossary References

1. INTRODUCTION Nanotechnology is a leading interdisciplinary science that is emerging as a distinctive field of research. Its advances and applications will result in technical capabilities that will allow

ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

the development of novel nanomaterials with applications that will revolutionize the industry in many areas [1, 2]. It is now well established that dimensionality plays a critical role in determining the properties of materials and its study has produced important results in chemistry and physics [3]. Nanoparticles are one of the cornerstones of nanotechnology. Indeed, even though the research in this field has been underway for a long time, many present and future applications are based on nanoparticles. For instance, the electron tunneling through quantum dots has led to the possibility of fabricating single-electron transistors [4–9]. One concept particularly appealing is a new three-dimensional periodic table based on the possibility of generating artificial atoms from clusters of all of the elements [10]. This idea is based on the fact that several properties of nanoparticles show large fluctuations, which can be interpreted as electronic or shell-closing properties with the appearance of magic numbers. Therefore, it is conceivable to tailor artificial superatoms with given properties by controlling the number of shells on a nanoparticle.

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (237–282)

238

Crystallography and Shape of Nanoparticles and Clusters

The development of nanotechnology can be approached from several directions; mesoscopic physics, microelectronics, materials nanotechnology, and cluster science. The different fields are now coming together, and a completely new area is emerging [11, 12]. Figure 1 illustrates how the different approaches are converging; it exhibits the domains of clusters and nanoparticles with different structures that result from an increase in the number of atoms. The different possible structures include nanorods, nanoparticles, fullerenes, nanotubes, and layered materials. One of the most remarkable advances in this field has been the synthesis of ligand-capped metallic clusters. In a seminal contribution, Brust et al. [13] used the classical two-phase Faraday colloid separation combined with contemporary phase-transfer chemistry to produce small gold nanoclusters coated with alkanethiolate monolayers. Several groups have pursued this technique [14–21], and have introduced improvements and modifications to the original technique. Whetten and his group [15, 16] made important contributions to this technique. A significant property of ligand-capped clusters is that they can be repeatedly isolated from and redissolved in common organic solvents without irreversible aggregation or decomposition. The properties of the monolayer-protected nanoparticles (MPNs and MPANs) allow handling in ways that are familiar to the molecular chemist since they are stable in air conditions. The MPNs of some metals such as Pt [22], Ag [23], Rh, and Pd [24] have been synthesized.

2. GEOMETRIC CONSIDERATIONS 2.1. Clusters and Nanoparticles Since the pioneering work of Ino and Ogawa [25, 26], it was clear that, in most cases, the structure of nanoparticles cannot be described by the bulk crystallography of the used material. The concept of multiple twins was used to explain many of the structures, Such as the icosahedron and the decahedron. This concept was directly imported from the macroscopic metallurgical studies, and was certainly very useful for a first understanding of the structure of nanoparticles. Another related research field was the study of clusters formed by few atoms. From that field, we learned that structures are formed by shells, and the concept of a magic number was introduced [27]. Nanoparticles referred

Clusters

Bulk

Molecules

1

Particles

2

10

10

10

1 0

103

104

105 100

1

10

106

The number of atoms Radius (A) ° (the number of inside atoms) (the number of surface atoms)

Figure 1. Domains of nanoparticles and clusters with different structures.

to sizes of ∼5–10 nm, and clusters referred to sizes of ∼1 nm. In recent times, the computational tools for study clusters allow the analysis of a large number of atoms, and the methods to study nanoparticles allow interrogating smaller nanoparticles. The two fields are merging into one. In this chapter we will use the term particle and cluster indistinctly.

2.2. Clusters with Cubic Symmetry When considering atom clusters of nanometric dimensions, it can be supposed that they correspond to some symmetry group. In previous works, small clusters (up to tents of atoms) of diverse forms (tetrahedron, hexahedron, octahedron, decahedron, dodecaheltahedron, trigonal, trigonal prism, and hexagonal antiprism, with and without a central site) have already been studied [28]. Partially and totally capped clusters were also considered (clusters to which a site has been added to each face of the polyhedron, equidistant to every site of the face), in order to vary the number of sites for all of the polyhedrons, and thus allowing the comparison as a function of the number of sites. Geometric characteristics of the clusters formed by concentric layers can be considered, as formed by equivalent sites: sites located at the same distance from the origin, which occupy the same geometric place and have the same environment, that is, the same number and type of neighbors. These layers can be arranged in such a way that they group in shells forming clusters of different sizes, retaining the original geometric structure. The number of shells in the cluster is called the order of the cluster, and is represented by letter . The studied structures were: the icosahedron, the face-centered cubic structure fcc (the cubooctahedron), the body-centered cubic bcc, and the simple cubic sc [29]. In order to determine the stability of the structures from an energetic point of view, a study of the cubo-octahedral and icosahedral structure was performed using the embedded atom method (EAM) for the transition metals Cu, Pd, Ag, and Ni [30]. It was determined that for sizes smaller than 2000 atoms, the icosahedron is the most stable structure, and for larger sizes, the cubooctahedron. For this study, the advantage of having equivalent sites in the clusters was used to reduce computation time. In a latter study, based upon the same metals, small clusters of less than 100 atoms with regular polyhedron geometries were used, and to change the cluster size, partially and totally capped clusters were considered. Tetrahedral clusters showed the highest stability for sizes of less than 18 atoms, and icosahedra for larger sizes [31]. In this work, the structural stability competition among different regular structures of the concentric shell type is searched. Here, the studied geometries are reproduced, considering some other arrays of the layers in shells, which give rise to other geometries. The fcc structure considers many structures, which are divided into two groups: (1) with a central site, and (2) without a central site. Among the centered ones, the cubooctahedron, octahedron, and truncated octahedron are considered. For the ones not centered, the

239

Crystallography and Shape of Nanoparticles and Clusters

octahedron and the truncated octahedron are considered. The bcc structure has the dodecahedron. The sc structure does not allow a study like the one presented here because, when a truncation is tried, second neighbor distances are obtained. Once the geometric properties of the structures are determined, the EAM is applied to them, and their stability is discussed.

(a)

(b)

(c)

2.3. Geometric Characterization The following procedure was used to determine the geometric characteristics of the structures: (1) First, identify the structures, that is, their geometric shape; (2) the nature of the site coordinates which conform the structures; and (3) their neighbors with surrounding layers, in order to identify the equivalent sites and generate the concentric shell-type structures. Once this previous stage is completed, the geometric properties to be determined are defined, the structures are presented, and the geometric properties of each structure are numbered. Afterwards, based on the geometric characteristics of each structure, the analytic expressions are deducted as a function only of the cluster order for the defined properties. An example of the concentric shell-type structures is shown in Figure 2, which presents the cubooctahedron in three different sizes. In this figure, a central site surrounded by 12 sites forming a cubooctahedron with one first shell can be seen; Figure 2(a). Then it is covered with another shell of sites distributed in three layers of equivalent sites, as will be seen later, but retaining the original geometric shape; Figure 2(b). And finally, Figure 2(c) presents a three-shell cubooctahedron. The structures considered are the face-centered cubic fcc, [Fig. 3(a)], and the body-centered cubic bcc [Figs. 3(d) and (e)]. Although both are cubic, in Figure 3(a) the fcc does not seem to be cubic, but the face-centered cubic structure with a central site fccc [Fig. 3(b)], and the face-centered cubic structure without a central site fccs [Fig. 3(c)], show that they do have a cubic shape: 8 vertexes, 6 square faces, and 12 edges. The bcc structure has a central site, while the fcc structure can be considered with and without a central site. Truncating each one of them in a certain direction, noncubic structures are obtained, they have, besides vertexes (V ), edges (E), and square faces (S), faces of different shapes: a triangular face (T ), a hexagonal face (H), and a rombohedral face (R). The resulting structures are: the

(d)

Figure 3. Structures (a) face-centered cubic fcc, (b) face-centered cubic with central site fccc, (c) face-centered cubic without central site fccs, and (d), (e) body-centered cubic bcc.

cubooctahedron [Fig. 4(a)], the octahedron [Fig. 4(b)], the truncated octahedron [Fig. 4(c)], and the dodecahedron [Fig. 4(d)]. For the fccc, the obtained structures are the cubooctahedron CO, the octahedron OCTAC, and the truncated octahedron OCTTC; for the fccs, the structures obtained are the octahedron OCTAS, and the truncated octahedron OCTTS; and for the bcc, only the decahedron is obtained, DODE. Table 1 presents the number of characteristic sites of each structure, in such a way that each structure is read as: the structure is formed by V vertexes attached by E edges forming X faces of different types, where X = S, T , H , R, as it corresponds. The icosahedron is included [Fig. 4(e)], because it can be obtained by an adequate distortion of the CO, and it is useful as a reference for a comparison of results. ICO, the fccc and its derivates, as well as the bcc and the DODE, present a central site. The OCTTC is centered over an octahedron with a central site of 19 sites, and the OCTTS over a regular octahedron of 6 sites.

2.3.1. Standard Coordinates Each of the structures studied here, except the ICO, corresponds to a well-defined crystalline structure, fcc or bcc, so that the coordinates of the geometric sites which compose them are characteristic of each one of them. In each structure, the standard coordinates are described by triads (a, b, c), a, b and c are integers, where the total number of sites is obtained by doing permutations and commutations

(a)

(a)

(b)

(c)

Figure 2. Cubooctahedrons formed by a central site and (a) one layer, (b) two layers, and (c) three layers.

(e)

(b)

(c)

(d)

(e)

Figure 4. Polyhedrons resulting from truncation of fcc and bcc structures. (a) cubooctahedron, (b) octahedron, (c) truncated octahedron, and (d) dodecahedron. (e) as a result of an adequate distortion of the cubooctahedron, the icosahedron is obtained.

240

Crystallography and Shape of Nanoparticles and Clusters

of the positive and negative values of a, b and c. Table 2 presents the characteristics of standard coordinates for each structure as a function of the cluster order . Sites ES [EH] in OCTTC are the sites on edges between hexagonal and squared [hexagonal] faces.

Table 1. Number of geometric sites corresponding to the structures of the first column. Last column presents the number of central sites of the structure.

ICO fccc CO OCTAC OCTTC fccs OCTAS OCTTS bcc DODE

V

E

S

T

H

R

Central site

12 8 12 6 24 8 6 24 8 14

30 12 24 12 36 12 12 36 12 24

— 6 6 — 6 6 — 6 6 —

20 — 8 8 — — 8 — — —

— — — — 8 — — 8 — —

— — — — — — — — — 12

1 1 1 1 19 — — 6 1 1

2.3.2. First Neighbors There will be different types of equivalent sites, depending on the geometric structure, and each type will occupy a geometric position which will be of type vertex (V ), edge (E), square face (S), triangular face (T ), hexagonal face (H), and rombohedral faces (R), and it will have a total number of neighbors, coordination z, which will correspond to the structure, but the number of neighbors with layers of its same shell, with interior and exterior shells, will be characteristic of each site. Table 3 presents the different types of sites in the studied structures, as well as the number of first neighbors with layers of interior shells (↓), in the same shell (↔), and with exterior shells (↑). Also, total coordination z

Table 2. Standard coordinates characteristics for each of the structures.

fccc

CO

OCTAC

OCTTC

fccs

OCTAS

OCTTS

bcc

DODE

Site

a

S E V

[2 − 1 2]

S T E V T E V S



2 +2

0

0  − 2

H

 − 2  + 1

/2 + 1   even

 + 1/2   odd

ES EH V

+2

S E V T E V S

c

b+c

0 2 − 1 [2 − 1 2]

0 2 − 2 (even) 2

0  − 2

 − 2

1  

1 

 0



1 2 − 2

1 2 − 1

0 0

0  − 2/2  even

0  − 3/2  odd

1 /2 + 2  even

1  + 3/2  odd

1 

+2

0 0



2 − 2 2 − 1

2 − 2 2 − 1

0 2 − 2

1 2 − 1 (odd) 2 − 1

1  − 1

1 2 − 1 2 − 1 +1

+1

V

+1

S E



0

0  − 2

0 0

0 /2  even

0  − 1/2  odd

1 

/2 

 − 1/2  

1 /2

1  − 1/2 0

0  − 2

0  − 2 (even)

1  − 2 (odd)

 

1  − 1

0 2/ − 2 2 − 1

0

0 0

a+b+c even [2 6 − 2] [2 6 − 2] 6 even

 2 − 2 2 2 2 2 2 2

 + 2 2 2 + 2



+1

H E

V R E V

b

2 + 2 2 + 2 2 + 2 odd

2 − 1 6 − 5

4 − 3 6 − 3 6 − 3 2 − 1 2 − 1 2 − 1

 − 4  − 2 2 + 1 2 + 1 2 + 1 even,  even odd,  odd

 3 − 4

2 3 − 2

2 + 1 3 − 2 3 2 − 1 2 − 1 2 − 1

241

Crystallography and Shape of Nanoparticles and Clusters Table 3. Number of first neighbors with shell on external layers ↑ (internal ↓) and with layers in the same shell ↔ for the different types of sites of the structures fcc, icosahedral, and bcc. The total coordination z is given by ↑ + ↓ + ↔.

↑ ↔ ↓ 4 4 4 fcc

CO

OCTAC

OCTTC

4 4 4 fccs

OCTAS

OCTTS

ICO

bcc

DODE

Faces T ↑ △ ↓

S ↑  ↓

H ↑ H ↓

R ↑ ♦ ↓

A ↑ \ ↓

V ↑ • ↓

— —

— — —

— — —

7 4 1

9 3 0

12

4 4 4

3 6 3

— — —

— — —

5 5 2

7 4 1

12

— — —

3 6 3

— — —

— — —

5 6 1

8 4 0

12

4 4 4

— — —

3 6 3

— — —

6 5 1

12

— —

— — —

— — —

7 4 1

9 3 0

12

— — —

3 6 3

— — —

— — —

5 6 1

8 4 0

12

4 4 4

— — —

3 6 3

— — —

5 5 2

6 5 1

12

— — —

3 6 3

— — —

— — —

4 6 2

6 5 1

12

4 0 4

— — —

— — —

— — —

6 0 2

7 0 1

8

— — —

— — —

— — —

2 4 2

3 4 1

4 5 3

4 5 3

4 6 2

4 6 2

5 5 2

5 6 1

4 4 0

z (a)

4 3 1

(b)

(c)

(d)

Figure 5. fcc polyhedrons of (a) 13 sites (central site plus a sublayer of 12 sites), (b) 63 sites (a complete layer of 62 sites), (c) 171 sites (a sublayer of 108 sites over the order 1 fccc), and (d) 365 sites (two complete layers).

1 × 12 E sites, 5 × 6 S and 8 V , but looking carefully it will be noticed that the S sites are not equivalent because the one in the center of the squared face has a coordination of 4 with the surface sites, while the other has a coordination of 6, but both have a coordination of 4 toward external layers. It is also observed that in Figure 5(c) the four sites of type S at the center of the face have a coordination of 4 with the surface sites, and the other eight sites of type S have a coordination of 5 with surface layers, but all of them have 4 toward external layers, while in Figure 5(d) it is observed that the S sites near the vertexes have a coordination of 6 with sites over the same surface shell, and the ones close to sites E have a coordination of 5, this is why in Table 3 the coordination for S sites can be of three types, but all of them with a coordination of 4 toward external layers. That is why in fccc and fccs there are shells composed of two subshells, one that will contain the surface sites, and the other that will remain internal. Surface sites will interact with the internal sites, and these internal ones with those of the cluster and the surface ones. From Fig. 4d it is worth noticed that the vertex sites for the dodecahedron are of two types: one on which four edges converge and another on which three edges converge. This is why in Table 3 the coordination for the V sites for the DODE structure are of two types.

8

in the structure is presented. The truncated octahedron with a central site presents two types of edges, the edges between squared and hexagonal faces, ES, with coordination (2, 5, 5), and the edges between hexagonal faces, EH, with coordination (1, 6, 5), while all of the other structures have only one. It should be noticed that, in order to obtain structures of the concentric shell type in the fccc and fccs structures, two shells will be needed, as shown on Figure 5 for fccc (similarly for fccs); Figure 5(a) presents the fccc “cube” of 13 sites; Figure 5(b) shows the fccc of 63 sites, which will be a complete shell of 62 sites (12 + 50) over the central site; Figures 5(c) and (d) present the fccc of 171 sites and the fccc of 365 sites, the latter is the one with two shells, one with 108 sites and the other with 194 sites. The surface sites in the fccc of 13 sites become internal when adding 50 to complete the fccc of 63 sites, the same happens for that one of the 171 sites when covering it to complete the one with 365. From Figure 5(b), it is observed that there will exist

2.3.3. Definition of Geometric Properties Table 4 presents definitions and their mathematical expressions for the general geometric characteristics, which can be expressed analytically as a function of only the order of the cluster , for each structure or geometric array. The upper section of Table 4 refers to the geometric characteristics of the cluster, the middle section to the bonds, and it relates Table 3, and the superior part of Table 4, and the lower section of Table 4 to the average coordination based on the rest of Table 4. The cluster can be defined as a nucleus and a surface over the nucleus. The number of I sites, the number of types of I sites and what refers to dispersions is over the surface, and the average number of bonds and coordination is over the involved sites of the cluster. In the expressions for CC and SS , ↑I means the number of I site bonds in the internal layer (the nucleus surface) with sites in the external layer (the cluster surface), ↓I means the number of I-site bonds in the external layer (the cluster surface) with sites in the internal layer (the nucleus surface), and ↔I means the number of I-site bonds in the surface of the cluster with

242

Crystallography and Shape of Nanoparticles and Clusters Table 4. General definitions of geometric parameters for different structures. Symbol

Definition

Mathematical expression

N N  NI RaI  Nshell T Nshell I CO CC SS C

Total number of atoms in a cluster of order  Number of atoms in a crust or on the surface Dispersion; Number of surface atoms on site I I = S T H R E V Number of shells in which are contained the I-sites (a = even, odd) Number of shells at the surface Total number of shells in the cluster Dispersion due to surface I sites; Number of internal bonds in a cluster of order ; Number of bonds between two adjacent crusts  − 1 and ; Number of bonds on the surface of a cluster of order ; Number of bonds in a cluster of order ;

Z CO Z CS Z SC Z SS Z

cluster average coordination; Core average coordination; Core–surface average coordination; Surface–core average coordination; Surface average coordination;

sites in the same surface. It must be clear that, due to the fact that there are many ways to coordinate S sites in the surface for fccc and  fccs structures, the formula for CC must only be used as I ↑I × NI  − 1, and only for these two structures, the proposed equity does not apply.

N /N



NI /N × z × N  − 1  I ↓I × NI  I ↑I × NI  − 1 = 1  I ↔I × NI  2 CO + CS + SS 1 2

2C /N  2CO /N  − 1 CC /N  − 1 CC /N  2SS /N 

An fcc structure is that in which the unit cell is a cube with sites in the vertexes and in the center of squared faces, this means that there are 14 sites on a cube, 8 vertexes, and 6 squared faces [Fig. 3(a)], also, it can be seen as a completely capped octahedron. When attaching many of these arrays with common or shared sites in faces and vertexes, clusters are obtained with sites in vertexes, edges, and squared faces. The origin of coordinates can be chosen to be at the center of the cube or on a vertex, obtaining structures with and without a central site, respectively.

in both columns corresponding to subshells, the first one is for the internal, and the second one is for the external or surface. N represents the number of sites added to the cluster of order  − 1, Nsup is the number of sites forming the surface, and it is the same for either of the considered cases because the internal shell does not form part of the surface. Besides this, the distinction is made among the difj ferent types of sites S forming the surface, NS , where j is the coordination of the S sites with sites in the same layer, j = 4 5 6. In the calculus of the number of bonds between the two sublayers, the coordination from inside to outside was considered, because all of the types of S sites have a coordination of 4 with the exterior shell. When grouping the layers of equivalent sites in shells in a different manner than the one given in fccc, geometries that present, besides squared faces, triangular and hexagonal faces will rise. Here, the cubooctahedron, octahedron, and truncated octahedron are considered.

2.4.1. fcc Structure with a Central Site

2.4.2. Cubo-octahedra

In the fcc structure with a central site, the cubic cluster of order 1 is a cube with 63 sites, distributed in 5 layers around a central site [Fig. 5(b)]; it can be seen that it is an array of 6 unit cells with common sites. The order 2 cluster is obtained when covering the one of order 1 with 302 sites distributed in 13 layers to obtain a total of 365 sites [Fig. 5(d)], and so on. Layers of equivalent sites can be identified (Table 1): squared face (S), edges (E) and vertexes (V ). Table 2 presents the standard coordinates and its characteristics, it can be seen that the sum of the standard coordinates a, b and c is an even number, and in Table 3 the coordination of each site. From Figures 5(a)–(d) it can be seen that a shell is composed of two subshells. Table 5 lists the geometric characteristics of the fccc structure, separating each shell into subshells to allow the calculus of surface dispersions and the number of bonds as defined in Table 4, and for each shell. General expressions for the geometric characteristics of Table 4 are presented in Table 6. These are shown when they are being considered as shells or subshells,

The cubo-octahedron can be obtained by truncating an fcc structure in direction (111), this is shown in Figure 6(b), where a cubo-octahedron of 55 sites is presented, and is the result of removing the 8 corners of a cube of 63 sites. It is

2.4. fcc Structure

Table 5. Geometrical characteristics for the face-centered cubic and with central site structure, fccc. 

NS

NE

NV

RS

RE

RV

 Nshell

T Nshell

N

1

0 30 30

12 12 24

0 8 8

0 2 2

1 1 2

0 1 1

1 4 5

2 6 6

12 50 62

13 63 63

2

72 150 222

36 36 72

0 8 8

2 6 8

2 2 4

0 1 1

4 9 13

10 19 19

108 194 302

171 365 365

3

240 366 606

60 60 120

0 8 8

6 12 18

3 3 6

0 1 1

9 16 25

28 44 44

300 434 734

665 1099 1099

N

243

Crystallography and Shape of Nanoparticles and Clusters Table 6. Particular expressions of the geometric characteristics for the face-centered cubic with a central site structure, fccc. Layer

Sublayers 32 3 + 24 2 + 6 + 1

N 2

NS6 R6S

96 − 48 + 14 48 2 + 2 48 2 + 2 3 32 + 24 2 + 6 + 1 24 1

48 2 − 48 + 12 48 2 − 48 + 12 48 2 − 48 + 12 32 3 + 24 2 + 6 + 1 — —

48 2 + 2 48 2 + 2 48 2 + 2 3 32 + 24 2 + 6 + 1 24 1

NS5 R5S

48 − 1 −1

— —

48 − 1 −1

NS4 R4S

24 212 + 2 − 22  2

— —

24 2 − 12 + 2 − 22  2

NE p RE

242 − 1 2

NV RV CO

8 1

N Nsup 

122 − 1  0 0 632 3 − 24 2 + 6 − 13

CC SS

1216 2 − 10 + 1 484 + 48

C

448 3 + 60 2 − 9 − 14

formed by 8 triangular faces and 6 squared ones, attached by 24 edges and 12 vertexes; see Table 1. Consequently, the surface sites are localized in squared faces (S), triangular faces (T ), edges (E), and vertexes). The CO of order 1 has a central site and a first shell with one single layer of 12 vertexes, Figure 6(a). The CO of order 2 is formed by adding one shell formed by 42 sites in order to complete 55 [Fig. 6(b)], distributed in three layers: one layer of 6 S sites, another of 24 E sites and a third one of 12 V sites. When adding a third shell of 92 sites, the third order CO of 147 sites is completed [Fig. 6(c)], sites are distributed in four layers: one of 24 S sites, another of 8 T sites, one more of 48 E sites, and a fourth one of 12 V sites. Successively, complete shells are added this way, forming clusters of order . The standard coordinates for the cubo-octahedron are listed in Table 2, the number of neighbors for each site can be found in Table 3, and the geometric characteristics of the CO are listed in Table 7; from this table, it is possible to obtain the general expressions for the geometric characteristics of the cubooctahedron defined in Table 4. These expressions are presented in Table 11, for the cubo-octahedron and all the other fccc structures to be studied.

(a)

(b)

(c)

(d)

(e)

Figure 6. Cubo-octahedron of (a) order 1 of 13 sites, (b) order 2 of 55 sites, (c) order 3 of 147 sites, and (d) order 6 of 923 sites. Observe the distortion of CO (d) to obtain (e) an icosahedron of order 6 of 923 sites.

8 1

Figure 6(e) presents an icosahedron of 923 sites of order 6. When comparing with the CO, which has the same number of sites, from Figure 6(d), the distortion made to obtain the ICO can be observed. Table 3 shows the number of neighbors for each site, and Table 8 reproduces the geometric characteristics corresponding to the icosahedron, which have been reported previously [29]. A great similarity can be noticed between Tables 7 and 8, in the columns referring to the types of sites or layers by shell (columns for T , E, V ), and in the sites at the surface, in the total (the two last columns of tables N and N ), and in the expressions for both in Table 11. Based on these latter, it is possible to obtain the dispersions due to the surface sites of both structures obtaining the graphic in Figure 7. Surface and vertex dispersion is exactly the same for both structures, they start close to 1 and in 1, respectively, and then they decrease. Dispersion in triangular faces, for both structures, shows a very similar behavior, they start in the same value, and increase with the size of the cluster, but DT in the ICO increases Table 7. Geometrical characteristics for the cubo-octahedra. 

NS

NT

NE

1 2 3 4 5 6 7 8 9 10

0 6 24 54 96 150 216 294 384 486

0 0 8 24 48 80 120 168 224 288

0 24 48 72 96 120 144 168 192 216

 T NV RS RT RE RV Nshell Nshell

12 0 12 1 12 1 12 3 12 3 12 6 12 6 12 10 12 10 12 15

0 0 1 1 2 3 4 5 7 8

0 1 1 2 2 3 3 4 4 5

1 1 1 1 1 1 1 1 1 1

1 3 4 7 8 13 14 20 22 29

1 4 8 15 23 36 50 70 92 121

N

N

12 42 92 162 252 362 492 642 812 1002

13 55 147 309 561 923 1415 2057 2869 3871

244

Crystallography and Shape of Nanoparticles and Clusters

Table 8. Geometrical characteristics for the icosahedra. 

NT

NE

NV

RT

RE

RV

 Nshell

T Nshell

N

N

1 2 3 4 5 6 7 8 9 10

0 0 20 60 120 200 300 420 560 720

0 30 60 90 120 150 180 210 240 270

12 12 12 12 12 12 12 12 12 12

0 0 1 1 2 3 4 5 7 8

0 1 1 2 2 3 3 4 4 5

1 1 1 1 1 1 1 1 1 1

1 2 3 4 5 7 8 10 12 14

1 3 6 10 15 22 30 40 52 66

12 42 92 162 252 362 492 642 812 1002

13 55 147 309 561 923 1415 2057 2869 3871

faster. Similarly for the edges, they increase rapidly from zero to a maximum value, and decrease slowly almost parallel. The predominance of DT and DC for large sizes has to be noticed for ICO and CO, respectively.

2.4.3. Octahedron with a Central Site The octahedron with a central site has 6 vertexes (V ), 12 edges (E), and 8 triangular faces (T ), Figure 8. In this figure, the sides of the triangular faces have a length of 2dNN , an even number times the distance to first neighbors; this means that the triangular faces do not present a central site, and the number of sites for each edge is odd. The number of neighbors for each site with sites in other layers can be found in Table 3, the standard coordinates in Table 2, and the geometric characteristics for the octahedron with a central site are listed in Table 9. From this latter, the general expressions of the geometric characteristics for the octahedron with a central site (OCTAC) can be deducted and are presented in Table 11. Figure 8(c) shows an octahedron of 891 sites with a central site. With the same method applied for the CO and the ICO, the dispersions due to surface sites can be calculated from Table 5, and the results are presented in the graphic of Figure 9. It can be observed that the surface dispersion 1

Dispersion

Squared faces (C) Triangular faces (C) Edges (C) Vertices (C e I) Surface (C e I) Triangular faces (I) Edges (I)

(a)

(b)

(c)

Figure 8. Octahedron of 19 atoms with central site. Smaller atoms are presented in (a) to allow observation of first neighbor bonds and central site in (b). (c) Octahedron of 891 atoms, order 5, and with a central site.

and dispersions due to edges and vertexes decrease, while the one due to the triangular faces starts from zero and increases rapidly, being predominant for larger sizes.

2.4.4. Truncated Octahedron with a Central Site A truncated octahedron is a structure centered over an octahedron of 18 sites (2 layers) with a central site (layer 0), and has 24 vertexes (V ) 36 edges (E), 6 squared faces (S), and 8 hexagonal faces (H ); Figure 10. Edges are of two types: 24 between squared and hexagonal faces (ES), and 12 between hexagonal faces (EH); the number of EH sites NEH is always equal to 12, one site in each edge (EH), while the one in ES sites NES depends on the order of the cluster, this means that the length of the edges ES is variable, and the one of EH edges is 2dNN . And so, the hexagonal faces are irregular, except when the number of ES sites is 24 and the number of EH sites is 12, which occurs in the cluster of order 2. The number of first neighbors for each site can be found in Table 3, standard coordinates are in Table 2, and Table 10 presents the geometric characteristics of each structure. Figure 10(c) shows the truncated octahedron of 711 sites with a central site. From Table 10 it can be observed that the number of EH sites is 12 and of V sites is 24 for all , besides this, it is possible to obtain the general expressions of the geometric characteristics for the truncated octahedron with a central site, as presented in Table 11. Table 9. Geometrical characteristics for the octahedra with a central site.

0.5

0 2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

N(1/3) Figure 7. Dispersions due to surface sites of the cubooctahedron (C) and icosahedron (I) as a function of the cubic root of the number of sites.



NT

NE

NV

RT

RE

RV

 Nshell

T Nshell

N

N

1 2 3 4 5 6 7 8 9 10

0 24 80 168 288 440 624 840 1088 1368

12 36 60 84 108 132 156 180 204 228

6 6 6 6 6 6 6 6 6 6

0 1 3 5 8 12 16 21 27 33

1 2 3 4 5 6 7 8 9 10

1 1 1 1 1 1 1 1 1 1

2 4 7 10 14 19 24 30 37 44

2 6 13 23 37 56 80 110 147 191

18 66 146 258 402 578 786 1026 1298 1602

19 85 231 489 891 1469 2255 3281 4579 6181

245

Crystallography and Shape of Nanoparticles and Clusters

2.4.5. EAM Applied to fcc Structures

Dispersion

1

Triangular faces Edges Vertices Surface

0.5

0

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

N(1/3)

Figure 9. Dispersions due to surface sites of the octahedron with a central site.

Figure 11 presents the dispersions for the truncated octahedron and with a central site, derived from the data of Table 10. The predominance of the dispersion due to the sites in hexagonal and squared faces for large sizes can be observed. Dispersions due to other sites and the surface one decrease with the size of the cluster. With support from Tables 4 and 6–10, it is possible to obtain the characteristics defined in the middle and lower parts of Table 4 for each and every studied structure. Characteristics in the middle part of Table 4 are presented in Table 11, and Figures 12–15 present the obtained graphics of the dependence of the number of bonds at the surface SS , the number of bonds between the surface and the internal cluster SC , the number of bonds in the internal cluster CO , and of the total number of bonds in the cluster C , respectively, for all of the structures studied here, the curves for the ICO and CO practically coincide, so that just the curve of CO will be presented, except for CO . It is observed that for the structures with a central site, and taking as reference the CO and the ICO, the OCTAC has a larger number of surface bonds and the OCTTC has fewer. But for SC , the CO and ICO surpass the OCTAC and the OCTTC. For CO , the OCTAC and the OCTTC surpass the CO. For C , there are some crossings for smaller sizes, while for larger sizes the OCTAC and OCTTC surpass the CO.

The embedded atom method is applied, in the Foiles version [32] and with the parameters of copper, to clusters with central site fcc structures in order to calculate the cohesion energy per atom, determining in this way the stability of the clusters. The results are compared with the ones from the icosahedra and cubooctahedra previously reported [30]. Figure 16 presents the cohesive energy graphic as a function of N 1/3 for the octahedron and truncated octahedron clusters of fcc with central site type, for sizes from 300 to 3400 atoms. From Figure 16 it can be seen that the truncated octahedra (△) have a higher stability than the icosahedra (•) and cubooctahedra (), while the octahedra (⋄) have it until sizes larger than 900 atoms. In a previous work, it was found that the CO presented a higher stability than the ICO for larger sizes; when analyzing the number of bonds in the clusters, it is found that this condition is present when the number of bonds of CO surpasses the ICO ones. The competition among fcc structures with a central site and the icosahedron observed in the graphic of Figure 16 must have the same origin.

2.5. fcc Structure without a Central Site In the fcc structure without a central site, the cubic cluster of order 1 has the same unit cell as the fcc structure of Figure 5(a), a cube with 14 sites, 8 in vertexes and 6 in the center of the squared faces. The second order cluster has 172 sites [Fig. 3(c)], which results from covering the first-order cluster with one shell of 158 sites, distributed in 8 layers; the second-order cluster is an array of 9 cubes attached by faces and common sites. The next cubic cluster is formed by the second-order one and a shell of 494 sites in 18 layers covering it, and so on. Equivalent sites layers can be identified: Table 1: squared face (S), edges (E) and vertexes (V ). As in the fccc structure, each shell is composed of two subshells. Table 12 lists the geometric characteristics of the fccs structures, and the separation of subshells for each shell is done in order to allow the calculus of the number of bonds as defined in Table 2, and for each shell also. General expressions for the geometric characteristics of Table 12 are presented in Table 13. Table 2 presents the standard coordinates and their characteristics, it can be seen that the sum of the standard coordinates a, b and c is an odd number, and in Table 3 the coordination of each site is presented. When grouping the layers of equivalent sites in shells, in a different way than given for fccs, the geometries presented are given: besides the squared faces, triangular and hexagonal faces are included. Here, the cubooctahedron, octahedron, and truncated octahedron will be considered.

2.5.1. Octahedron without a Central Site

(a)

(b)

(b)

Figure 10. Truncated octahedron with central site of (a) order 1 and 79 atoms, (b) order 2 and 201 atoms, and (c) order 4 and 711 atoms.

The octahedron is formed by 6 vertexes, 12 edges, and 8 triangular faces. Figure 17 shows the octahedrons of 44 and 146 sites, without a central site. In these figures, it can be seen that there are sites in vertexes (V ), edges (E), and in triangular faces (T ). In this structure, the sides of the triangular faces have a length of 2 − 1dNN , an odd number times the distance to first neighbors dNN , this means that the triangular faces can present a central site, and the number

246

Crystallography and Shape of Nanoparticles and Clusters

Table 10. Geometric characteristics for the truncated octahedra with a central site. 

NS

NH

NES

NEH

NV

RS

RH

RES

REH

RV

 Nshell

T Nshell

N

N

1 2 3 4 5 6 7 8 9 10

0 6 24 54 96 150 216 294 384 486

24 56 96 144 200 264 336 416 504 600

0 24 48 72 96 120 144 168 192 216

12 12 12 12 12 12 12 12 12 12

24 24 24 24 24 24 24 24 24 24

0 1 1 3 3 6 6 10 10 15

1 3 3 4 6 7 9 11 13 15

0 1 1 2 2 3 3 4 4 5

1 1 1 1 1 1 1 1 1 1

1 1 1 1 1 1 1 1 1 1

3 6 7 11 13 18 20 27 29 37

5 11 18 30 43 61 81 108 137 174

60 122 204 306 428 570 732 914 1116 1338

79 201 405 711 1139 1709 2441 3355 4471 5809

of sites in each edge is even. The number of first neighbors is presented in Table 3, the standard coordinates in Table 2 and the geometric characteristics for the octahedron without a central site are listed in Table 14. Figure 17(c) shows the octahedron of 1156 sites without a central site.

The number of V sites NV is equal to 6 for all , and they are of just one type. From Table 14, it is possible to determine, for the octahedron without a central site, the dependence of the geometric characteristics on the order of the cluster , and it is reported in Table 14. This same

Table 11. Particular expressions of the geometric characteristics for the icosahedra and fcc with central site structures. CO

N 

Reven S Rodd S NT Rm T NH Rm H NE

OCTAC

OCTTC

16 3 14  + 8 2 +  + 1 3 3 16 2 − 48 + 2

10 3 107  + 21 2 +  + 19 3 3 10 2 + 32 + 18



3 2 − 48 + 2 16 3 + 24 2 + 14 + 3 —





310 2 + 32 + 18 10 3 + 63 2 + 107 + 57 6 − 12  + 2 8 2 − 1 8 4 2 + 20

10 3 11  + 5 2 +  + 1 3 3 10 2 + 2 30 2 + 6 10 3 + 15 2 + 11 + 3

N

NS

ICO

6 − 12  + 2 8 2 − 1 8 4 − 1 − 2 A — — 24 − 1





10 − 1 − 2 B — — 30 − 1

8 − 12 − 1 — — — 122 − 1

20 3 + 15 2 + 7

32 3 − 48 2 + 2811

4 2 + 20 C 24 − 1  2 −1 2 20 3 + 66 2 + 22 + 6

123 2 − 3/nu + 1 24 2

65 2 − 5/nu + 2 30 2

124 2 − 4 + 1 48 2

123 2 + 5 + 1 24 2 + 4 + 2

45 3 − 12 2 + 10 − 3

20 3 − 45 2 + 37 − 12

216 3 + 24 2 − 10 + 3

210 3 + 63 2 + 89 + 33

Reven E Rodd E CO

45 2 + 3 + 1

CC SS C

A=

 2 −1 2

 m/2−1 m/2      3" + a + 3"#     "=1 "=1   

     m−1/2    1+a   + 6" + a#  1 + a "=1

C=

−1 −1

m even 

 m/2−1 m/2 b       3" + a + 3" + 1#     "=1 "=1 1   

 = 3m + a# a = −1 0 1#

     m−1/2 b     1+a   + 6" + a + 1#  1 + a "=1 1

m odd

B=

 m/2−1 m/2      3" + a + 3"#     "=1 "=1   

     m−1/2    1+a   + 6" + a#  1 + a "=1

m even 

2 = 3m + a# a = −1 0 1# m odd

m even 

 = 3m + a# a = −1 0 1# m odd

where b is the integer of  + m + a + 1/2

247

Crystallography and Shape of Nanoparticles and Clusters 1 15 CO OCTAC OCTTC OCTAS OCTTS DODE

10

NSC(1/3)

Dispersion

Squared faces Hexagonal faces Edges ES Edges EH Vertices Surface

0.5

5

0

0 2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

N(1/3)

(1/3)

N

procedure is applied for the truncated octahedron, and without a central site and with the ones from the dodecahedron, which will be studied afterwards. The surface dispersion and the one due to each surface site also can be obtained from Table 14, and are presented in the graphic of Figure 18. The strong predominance of the sites of triangular faces and the fast decrease of the dispersions due to the vertexes and edges can be observed.

Figure 13. Number of bonds between the surface and the internal cluster as a function of the number of sites.

25 CO OCTAC OCTTC OCTAS OCTTS DODE

20

NCO(1/3)

Figure 11. Dispersions due to surface sites of truncated octahedron with central site.

15

10

2.5.2. Truncated Octahedron without a Central Site

5

The truncated octahedron without a central site, Figure 19, is formed by a central octahedron surrounded by 6 squared faces and 8 irregular hexagonal faces (3 sides of length 1 and 3 of length equal to the order of the cluster , in units of dNN , alternated) attached by 24 vertexes and 36 edges, so the edges are of two types: 24 between squared and hexagonal faces and 12 between hexagonal faces; the first ones are

0

2

3

4

5

6

7

8

9

10

11

12

14

15

16

Figure 14. Number of internal bonds as a function of the number of sites.

15

25 CO OCTAC OCTTC OCTAS OCTTS DODE

20

NC(1/3)

NSS(1/3)

10

15 CO ICO OCTAC OCTTC OCTAS OCTTS DODE

10 5

5

0

13

N(1/3)

2

3

4

5

6

7

8

9

N

10

11

12

13

14

15

16

(1/3)

Figure 12. Number of surface bonds as a function of the number of sites.

0

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

N(1/3)

Figure 15. Total numbers of bonds as a function of the number of sites.

248

Crystallography and Shape of Nanoparticles and Clusters

2.5.3. EAM Applied to fcc Structures without a Central Site

Cohesive energy/atom

3.45

The embedded atom method is applied, in the Foiles version [32] and with the parameters of copper, to clusters with fcc structures without a central site to calculate the cohesion energy per atom, determining in this way the stability of the clusters. The results are compared with the ones from the icosahedra and cubooctahedra previously reported [30]. Figure 21 presents the cohesive energy graphic as a function of N 1/3 for the octahedron and truncated octahedron clusters of fcc without central site type, for sizes from 300 to 3400 atoms. From Figure 21, it can be seen that the truncated octahedra (△) as well as the octahedra (⋄) compete in stability with the icosahedra (•) and cubooctahedra (). And most of all, the truncated octahedra show a higher stability than the octahedra.

3.35

Icosahedrons Cubooctahedrons Octahedrons with a central site Truncated octahedron with a central site 3.25

6

7

8

9

10

11

12

13

14

15

16

N(1/3)

Figure 16. Cohesive energy per atom as a function of the cubic root of the number of atoms in the cluster for the octahedrons and truncated octahedrons type fcc with a central site.

length dependent on the cluster order, and the second ones on length 1dNN , this means that there is not a single site in them, so the first 24 are simply called E. The number of first neighbors of each site is presented in Table 3, the standard coordinates in Table 2, and the geometric characteristics for the truncated octahedron without a central site are listed in Table 15. Figure 19(b) shows the truncated octahedron of 1288 without a central site. The number of V sites NV is equal to 24 for all , and they are of just one type. From Table 15 the expressions for the geometric characteristics for the truncated octahedron without a central site can be determined, and are presented in Table 16. Also, the surface dispersions can be determined, and are presented in the graphic of Figure 20. A not so strong predominance of the sites in squared and hexagonal faces can be observed, while the other dispersions decrease as the cluster size increases. With support from Tables 4 and 15 it is possible to obtain expressions for the number of bonds for the different types of bonds in the truncated octahedron without a central site, which are presented in Table 16 and whose graphics are shown in Figures 12–15.

Table 12. Geometric characteristics for the face-centered cubic without a central site structure, fccs. 

NS

NE

NV

RS

RE

RV

 Nshell

T Nshell

N

1

0 6 6

0 0 0

0 8 8

0 1 1

0 0 0

0 1 1

0 2 2

0 2 2

0 14 14

0 14 14

2

24 78 102

24 24 48

0 8 8

1 4 5

1 1 2

0 1 1

2 6 8

4 10 10

48 110 158

62 172 172

3

144 246 390

48 48 96

0 8 8

4 9 13

2 2 4

0 1 1

6 12 18

16 28 28

192 302 494

364 666 666

N

2.6. bcc Structure The body-centered cubic structure is a cube with sites in the vertexes and in the center of the cube, that is, there are 9 sites over a cube, 8 in vertexes, and 1 in the center of it; this is a cluster of order 1 with cubic arrays [Fig. 22(a)]. When attaching many of these arrays, clusters with sites in vertexes, edges, and squared faces are obtained. The cubic cluster of second order consists of an array of 8 cubes, which results from covering the first cube with a shell of sites distributed in three layers: 6 in squared face (S), 12 in edge (E), and 8 in vertexes (V ) [Fig. 22(b)]. The next cluster has 56 sites distributed in 3 layers, and so on. From here, it is deduced that it is possible to identify layers of equivalent sites; these layers can be grouped into shells yielding noncubic geometries which present sites in vertexes, edges, and in rombohedral faces, such as in the dodecahedron.

2.6.1. Dodecahedra The dodecahedron is formed by a central site, 14 vertexes of two types (V 1 and V 2), 24 edges (E) and 12 rombohedral faces (R), Figure 23. The two types of vertexes are: (1) 6 vertexes (V1 ) of coordination 4 with sites of their own shell and of4 with sites of exterior shells; an (2) 8 vertexes (V2 ) of coordination 1 towards sites of interior shells, 3 with sites of its own shell, and 4 with sites of exterior shells. The number of first neighbors is found in Table 3, standard coordinates in Table 2, and the geometric characteristics of the decahedron are listed in Table 17, from which the dependence of geometric characteristics for the decahedron can be determined and are presented in Table 16. Surface dispersions originating from the different types of sites in the surface of the decahedron can also be shown graphically; Figure 24. Based on Tables 3 and 17, it is possible to construct the expressions for the number of bonds for the different types of bonds in the dodecahedron, and are presented in Table 16 and their graphics in Figures 12–15.

2.6.2. EAM Applied to bcc Structures The embedded atom method is applied, in the Foiles version [32] and with the parameters of copper, to clusters with bcc structures to calculate the cohesion energy per atom,

249

Crystallography and Shape of Nanoparticles and Clusters Table 13. Particular expressions of the geometric characteristics for the face-centered cubic without a central site structure, fccs. Layer

Sublayers 32 3 − 24 2 + 6

N 2

NS6 R6S

96 − 144 + 62 48 2 − 48 + 14 48 2 − 48 + 14 32 3 + 24 2 + 6 + 1 24 1

48 − 12 48 − 12 48 − 12 3 32 + 24 2 + 6 + 1 — —

48 2 − 48 + 14 48 2 − 48 + 14 48 2 − 48 + 14 32 3 + 24 2 + 6 + 1 24 1

NS5 R5S

242 − 3 −1

— —

242 − 3 −1

NS4 R4S

6 2 − 22 + 2 − 32   − 1

— —

6 2 − 22 + 2 − 32   − 1

NE p RE

48 − 1 2 − 1

NV RV CO

8 1

N Nsup 

24 − 1 −1 0 0 1216 3 − 36 2 + 27 − 7

CC SS

248 2 − 13 + 5 248 2 − 6 + 1

C

1216 3 − 4 2 − 11 + 5

determining in this way the stability of the clusters. The results are compared with the ones from the icosahedra and cubooctahedra previously reported [30]. Figure 25 presents the cohesion energy graphic as a function of N 1/3 for dodecahedral clusters, for sizes from 300 to 3400 atoms. From Figure 25 it can be seen that the dodecahedra (⋄) compete in stability with the icosahedra (•) and cubooctahedra ().

3. CLUSTERS AND NANOPARTICLES WITH PENTAGONAL SYMMETRY 3.1. Introduction A cluster is defined as an aggregate of atoms; this can lead to clusters from two atoms (diatomic molecules), a lineal array of atoms, bidimensional, or three-dimensional arrays. This work presents the study of clusters with pentagonal symmetry, with sizes up to thousands of atoms in arrays of spherical or concentric layer types.

8 1

Arrays of linked atoms forming three-dimensional clusters are considered here as sites in geometric positions attached by the edges in such a way that faces of diverse forms are generated (triangular, squared, rombohedral, etc.). The distance between the sites is considered as the distance to first neighbors dNN , which is normalized to one. There could be sites in the vertexes, edges, and faces, either in the surface or internals; also, there could be different types of sites, depending on their position and the number and type of neighbors in the geometric array. There also could be equivalent sites, which present the same geometric characteristics: to the same distance from the center of the geometric array, in the same type of site and with the same number and type of neighbors. From the bicaped hexahedron or decahedron, Figure 26(a), pentagonal symmetry structures can be obtained. Among the clusters with structures of pentagonal symmetry, the following structures are considered: decahedra with and without a central site, icosahedra, pentadecahedra, Table 14. Geometric characteristics for the octahedra without a central site.

(a)

(b)

(c)

Figure 17. Octahedron without central site of (a) order 2 with 44 sites, (b) order 3 with 146 sites, and (c) order 6 with 1156 atoms.



NT

NE

NV

RT

RE

RV

 Nshell

T Nshell

N

N

1 2 3 4 5 6 7 8 9 10

0 8 48 120 224 360 528 728 960 1224

0 24 48 72 96 120 144 168 192 216

6 6 6 6 6 6 6 6 6 6

0 1 2 4 7 10 14 19 24 30

0 1 2 3 4 5 6 7 8 9

1 1 1 1 1 1 1 1 1 1

1 3 5 8 12 16 21 27 33 40

1 4 9 17 29 45 66 93 126 166

6 38 102 198 326 486 678 902 1158 1446

6 44 146 344 670 1156 1834 2736 3894 5340

250

Crystallography and Shape of Nanoparticles and Clusters

Dispersion

1

Table 15. Geometric characteristics for the truncated octahedra without a central site.

Triangular faces Edges Vertices Surface

0.5

0

2

3

4

5

6

7

8

9

10

11

12

13

14

15

NS

NH

NE

1 2 3 4 5 6 7 8 9 10

0 6 24 54 96 150 216 294 384 486

8 0 24 24 48 48 80 72 120 96 168 120 224 144 288 168 360 192 440 216

 T NV RS RH RE RV Nshell Nshell

24 0 24 1 24 1 24 3 24 3 24 6 24 6 24 10 24 10 24 15

1 1 2 3 4 5 7 8 10 12

0 1 1 2 2 3 3 4 4 5

1 1 1 1 1 1 1 1 1 1

2 4 5 9 10 15 17 23 25 33

3 7 12 21 31 46 63 86 111 144

N

N

32 78 144 230 336 462 608 774 960 1166

38 116 260 490 826 1288 1896 2670 3630 4796

16

N(1/3)

Figure 18. Dispersions caused by surface sites of octahedron without a central site as a function of the cluster size.

truncated decahedra (Marks decahedra), star-type decahedra, modified and developed decahedra, truncated icosahedra, and the decmon.

3.2. Decahedra Decahedra are obtained from the bicapped hexahedra, and also from attaching two pentagonal-based pyramids from their bases and sharing their sites (which form the equator of the cluster), yielding geometrical bodies of 7 vertexes (2 at poles and 5 at equator), 15 edges (all from the same length, 5 at the equator), and 10 triangular equilateral faces: 5 of them converge on each pole, and by pairs, they form the edges of the equator. Decahedra can be without a central site, Figures 26(a) and (c), and with a central site, Figure 26(b), without losing the decahedral form. So, decahedra have the vertexes at the poles VP, vertexes at the equator VE, at the edges over the equator EE, edges at poles EP, and in triangular faces T . It has to be noticed that coordination, that is, distribution to first neighbors (NN), is what makes the difference in each type of sites, although the total coordination is the same for all of the sites of the corresponding cluster. Table 18 presents the coordination of each site, for example, the poles (VP) have 4 first neighbors (NN)

(a)



(b)

Figure 19. Truncated octahedron without a central site of (a) order 3 with 260 sites and (b) order 6 with 1288 sites.

with sites at their same shell, 1 NN toward the inner shell, and 4 toward the exterior shell. The decahedron of order 1, without a central site, Figure 26(a), has only 7 vertexes in two layers; the one from order 2, Figure 26(c), is obtained from covering the one of order 1 with a shell of 47 sites distributed as follows: 7 V sites of two types, 30 E sites in three layers (10 sites of one type ant the equator), and 10 at triangular faces (sites T , one for each triangular face) in one single layer, for a total of 54 sites in the cluster. Decahedra of superior order are formed by coverage of this cluster of order 2 with successive shells of many layers in each one. The decahedron with central site of order 1, Figure 26(b), has 15 sites E, one per edge, of two types, 5 sites of one type at the equator, and 7 sites V , a total of 22 sites and the central one in five layers. The second-order cluster results from the order 1 cluster covered by a shell of 82 sites distributed in 8 layers: 45 E sites in 5 layers, 30 T sites in one single layer, and 7 V sites in two layers, for a total of 105 sites in the cluster, and so on for cluster of superior order. Table 19 presents the geometric characteristics of decahedra with and without a central site. The first column, common for all decahedra, lists the cluster order . This is followed by two groups of 9 columns each, which correspond to the decahedron with and without central site. The first three columns of each group list the number of sites on each type of site in the cluster: triangular face (T ) NT , edge (E) NE , and vertex (V ) NV . The two following columns show the number of sites per shell NSE , and the total of sites in the plane just above the equator NtSE , (the SE sites). The following two columns list the number of sites at the equator (sites EC) per shell NEC , and in the whole cluster, NtEC ; the joint between SE and EC sites form the SEC layer, which will be needed afterwards, and they both form a pentagonal flat layer. Finally, the two last columns of each group, the number of sites in the shell N , and the total of sites in the cluster N . Decahedra with (without) a central site have an odd (even) number of sites per edge, which is important for the star and truncated-type decahedra, detailed later. From Table 19 it is observed that, for both decahedron types, the number of vertexes per shell is 7, 2 type VP and 5 type VE; We can also obtain, for each type of decahedra, the dependence with the cluster order of the number of T sites, E sites, the number of sites in the layer at the equator, as well as the total sites in the shell (or of the surface), and from the total of sites over the equator, at the equator and in the decahedron. Such dependency is expressed in the

251

Crystallography and Shape of Nanoparticles and Clusters Table 16. Particular expressions of the geometric characteristics for the fcc without a central site and dodecahedral structures. OCTAS

OCTTS

16 3 2  +  3 3 2 16 − 16 + 6 316 2 − 16 + 6 16 3 + 2

47 10 3  + 13 2 +  + 6 3 3 10 2 + 16 + 6 310 2 + 16 + 6 10 3 + 39 2 + 47 + 18

NS



p S



RiS



NT Rm T NH Rm H NR

82 − 3 − 1 A — — —

6 − 12  + 2 8 2 − 1 — 8 — — 4 + 1 C —





N N 

R

p

RR

DODE

12 2 + 2 12 2 + 2 4 3 + 6 2 + 4 + 1 — — — — — — — 12 − 12 2 4 2 − 1 4 24 − 1

RiR





NE

24 − 1

44 3 − 6 2 + 4 − 7

RiE

−1

CO

48 3 − 24 2 + 25 + 7

24 − 1  2 −1 2 210 2 + 9 − 1

CC SS

48 − 12 124 2 − 4 + 1

65 2 − 5 + 2 24 + 2

24 − 1 46 2 + 1

C

48 3 − 11 + 40

210 2 + 39 + 29

82 3 + 3 2 −  − 3

p

−1

RE

following relations for the decahedra with a central site; it is to be noticed that, for  = 1, there are 11 sites at the equator because the central site is included:  NEP  = 102 − 1 NE  = 152 − 1 NEE  = 52 − 1 NT  = 102 2 − 1 − 1 1 Squared faces Hexagonal faces Edges Vertices Surface

−1 −1

NSE  = 10 − 1 + 5 NtSE  = 5 2 NEC  = 10 NtEC  = 5 + 1 + 1 N  = 20 2 + 2 N  =

Dispersion

4 3 + 6 2 + 4 + 1

16 20 3  + 10 2 +  + 1 3 3

and for decahedra without a central site,  NEP  = 20 − 1 NE  = 30 − 1 NEE  = 10 − 1

0.5

NT  = 20 2 − 50 + 30 NSE  = 10 − 1 NtSE  = 5 − 1 + 1 NEC  = 52 − 1

0

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

N(1/3)

Figure 20. Dispersions due to surface sites of the truncated octahedron without a central site as a function of the cubic root of the cluster size.

NtEC  = 5 2 N  = 20 2 − 20 + 7 N  =

20 3   + & 3 3

252

Crystallography and Shape of Nanoparticles and Clusters

Cohesive energy/atom

3.45

3.35

Icosahedron Cubooctahedron Octahedron without central site Truncated octahedron without central site 3.25

6

7

8

9

10

11

12

13

14

(a)

15

16

(b)

Figure 23. bcc dodecahedron of (a) order 1 with 15 sites, and (b) order 2 with 175 sites.

N(1/3)

Figure 21. Cohesion energy per atom as a function of the square root of the number of atoms in the cluster, for the octahedron and truncated octahedron with central site, type fcc.

Note that from the expression for NA , and from Figures 26(c) and (d), the number of EP and EE per edge is 2 − 1 [2 − 2] for decahedra with [without] a central site. If, instead of attaching the base of the two pentagonal base pyramids to form the decahedron, they are separated, two types of geometrical figures can be obtained, depending on the way the sites of the bases are connected: the pentadecahedron and the icosahedron.

3.3. Pentadecahedra If the pentagonal pyramids which generate the decahedron when they are joined, are separated, they are in a position in which the vertexes of the base of both can be attached by 5 edges (EC) perpendicular to the base, yielding 5 rectangular faces, which form the pentadecahedra, Figure 27, that is, a decahedron with a wide waist or a developed decahedron. Separation can be done by adding intermediate layers of type SEC (formed by a layer of sites SE and another of sites EC as mentioned previously). The number of SEC intermediate layers, and the length of the edges perpendicular to the bases, depend on the desired separation. For example, the pentadecahedron from Figure 27 has 5 layers in the waist; to obtain it, 4 layers of type SEC are added to the order 6 decahedron of 1442 sites without a central site, which results

in edges 4dNN long. It has to be noticed that, upon widening of the waist of a decahedron, from one equatorial layer to two equatorial layers, an SEC-type layer is added, when widening three equatorial layers, two SEC-type layers are added, that is, for each equatorial layer wanted to widen the decahedra waist, one type of SEC layer is added. The pentadecahedra are polyhedra of 12 vertexes (2 poles and 10 in vertexes at the waist), 25 edges (20 of one type and 5 of the other type, which join the vertexes of the two pyramids, and whose length depends on the number of intermediate layers added), 10 equilateral triangular faces and, 5 rectangular lateral faces (or squared, depending on the number of intermediate layers added) (Fig. 27). The number of sites in these pentadecahedra depends on the size of the original decahedron and how many intermediate layers are added; also, whether they are with and without a central site is considered, depending on the original decahedron from which they were generated. So, at the surface of the pentadecahedra, there are the same type and number of sites as in the decahedra, plus the NEC sites added (because the NSE are not at the surface), which are divided in sites type VE, sites EE, sites at vertical edges at the width of the waist EV and in rectangular faces RF. Table 20 presents the coordination of each type of site in the pentadecahedron. Notice that, upon comparison of Table 18, as expected, only the sites corresponding to the pentadecahedron are added, and those of the decahedron are not. Sites NSE added, originally 5 EP sites, and the rest T sites (except for the pentadecahedron of order 11, without a central site) are converted in internal sites of coordination 12. Table 17. Geometric characteristics for the dodecahedra.

(a)

(b)

Figure 22. bcc structure of (a) 9 sites, and (b) 35 sites.



NR

NE

NV

RR

RE

RV

 Nshell

T Nshell

1 2 3 4 5 6 7 8 9 10

0 12 48 108 192 300 432 588 768 972

0 24 48 72 96 120 144 168 192 216

14 14 14 14 14 14 14 14 14 14

0 1 2 4 6 9 12 16 20 25

0 1 2 3 4 5 6 7 8 9

2 2 2 2 2 2 2 2 2 2

2 4 6 9 12 16 20 25 30 36

2 6 12 21 33 49 69 94 124 160

N

N

14 50 110 194 302 434 590 770 974 1202

15 65 175 369 671 1105 1695 2465 3439 4641

253

Crystallography and Shape of Nanoparticles and Clusters

Dispersion

1

Rhombohedral faces Edges Vertices Surface

0.5

0

2

4

3

5

6

7

8

9

10

11

12

13

14

15

(a)

(b)

(c)

(d)

16

N(1/3)

Figure 24. Dispersion due to surface sites of the dodecahedron as a function of the cubic root of the cluster size.

For pentadecahedron order, we can use ',  for the decahedron order from which comes and ' for the number of layers at the waist, so regular decahedra would be pentadecahedra with ' = 1. The number of layers of type SEC which are added to the decahedron referred to generate the pentadecahedron is ' − 1, so that the number of sites added is equal to the sum of the total of sites SE and the total of EC sites of the corresponding decahedron, columns 6 and 8, and 15 and 17 of Table 19. So, in order to have a pentadecahedron of order 55 with a central site, one has to start with a decahedron of order 5, 111 sites, and an aditional 4 layers of type SEC of 125 + 151 = 276 sites each, for a total of 1104 added sites and 2215 sites in the cluster. The number of sites T , EP, and VP is the same as in the decahedron which originated from the pentadecahedron. The number of sites EE and VE is duplicated with respect to the original decahedron. The number of EV and RF sites

Cohesive energy/atom

3.45

Figure 26. Decahedral polyhedrons. (a) bilayered hexahedron or decahedron of 7 atoms of order 1, without central site. (b) Decahedron of 23 atoms of order 1 with central site. (c) Decahedron of 835 sites of order 5, without central site. (d) Decahedron of 1111 atoms of order 5 with central site.

for the pentadecahedra with and without a central site is the same, and is presented in Table 21, which shows the geometric characteristics of the pentadecahedra with and without a central site respectively; only some of ' values are presented. There are three groups of three, four, and five columns, respectively. In the columns of the first group are listed the quantities common to the two types of pentadecahedra: cluster order  and ', and the number of sites EV , NEV ; for both polyhedra, in each following group, the geometric characteristics for each polyhedron are presented, those of the pentadecahedron both with and without a central site. A list of the number of RF sites NRF , the number of sites added Nag , of sites at the surface N, and the total of sites N in the cluster of order ' is given. Note that, for ' = 1, the values from Table 19 are obtained. From Table 21, the analytic expressions for the number of sites EV can be obtained as a function of  and : NEV ' = 5' − 2 as well as the number of RF sites, sites which are added, sites in the surface, and the total of sites for pentadecahedra

3.35 Icosahedrons Cubooctahedrons Dodecahedrons (bcc)

Table 18. Coordination or number of first neighbors (NN) of the different types of sites in the decahedron with sites in external, the same, and internal shell. 3.25 6

7

8

9

10

11

12

13

14

15

16

N(1/3)

Figure 25. Cohesive energy per atom as a function of the cubic root of the number of atoms in the cluster for decahedron bcc type.

Shell

T

EE

EP

VE

VP

External Same Internal

3 6 3

6 6 0

4 6 2

8 4 0

6 5 1

254

Crystallography and Shape of Nanoparticles and Clusters

Table 19. Geometric characteristics for the decahedra with and without a central site.  is the order of the cluster, NI with I = T , E and V , is the number of sites I. The number of sites SE and EC in each layer, the number of sites in each layer, the total number of SE and EC sites and the total number of sites in the cluster are also listed. With a central site

Without a central site



NT

NA

NV

NSE

NtSE

NEC

NtEC

N

N

NT

NA

NV

NSE

NtSE

NEC

NtEC

N

N

1 2 3 4 5 6 7 8 9 10

0 30 100 210 360 550 780 1050 1360 1710

15 45 75 105 135 165 195 225 255 285

7 7 7 7 7 7 7 7 7 7

5 15 25 35 45 55 65 75 85 95

5 20 45 80 125 180 245 320 405 500

10 20 30 40 50 60 70 80 90 100

11 31 61 101 151 211 281 361 451 551

22 82 182 322 502 722 982 1282 1622 2002

23 105 287 609 1111 1833 2815 4097 5719 7721

0 10 60 150 280 450 660 910 1200 1530

0 30 60 90 120 150 180 210 240 270

7 7 7 7 7 7 7 7 7 7

1 10 20 30 40 50 60 70 80 90

1 11 31 61 101 151 211 281 361 451

5 15 25 35 45 55 65 75 85 95

5 20 45 80 125 180 245 320 405 500

7 47 127 247 407 607 847 1127 1447 1807

7 54 181 428 835 1442 2289 3416 4863 6670

with a central site NRF  ' = 52 − 1' − 2 Nag  ' = ' − 110 2 + 5 + 1 N  ' = 20 2 + 2 + 10' − 1 N  ' =

16 20 3  + 10 2 +  + 1 + Nag  ' 3 3

and for pentadecahedra without a central site NRF  ' = 10 − 1' − 2 Nag  ' = ' − 110 2 − 5 + 1 N  ' = 20 2 − 20 + 7 + 52 − 1' − 1 N  ' =

20 3   + + Nag  '& 3 3

3.4. Modified Decahedra In a certain decahedron, with or without a central site, the number of sites per edge increases in 2 sites per shell increased. For example, in the decahedron of 287 sites of

order 3 with a central site, Figure 28(a), in the surface shell there are 5 sites in each edge, and in the one of order 4, there are 609 sites, Figure 28(b), there are 7 sites per edge. The decahedron modification can be done in two ways: (1) as channels at the surface, and (2) as surface reconstruction. Besides, modified and developed decahedra can be obtained, adding SEC-type layers in order to widen the waist of the modified decahedra.

3.4.1. Decahedra with Additional Faceting (Channels at Twin Boundaries) When suppressing surface sites from the edges converging in the poles (EP sites) and from all of the vertexes in a decahedron, a one-channel decahedron is obtained, for example, from the decahedron of 609 sites of order 4 with a central site [Fig. 28(b)], a decahedron of 532 sites is obtained with one channel convergent to the pole, Figure 29(a), eliminating 77 sites. These result in a polyhedron formed by a decahedron of order 3 with triangular faces covered with sites corresponding to the decahedron of order 4, in a way that some surface sites that were T type are now type E. If, now, the following edges (which were type T sites) are eliminated from the surface, a triple-channel decahedron of 412 sites is obtained, 120 sites were eliminated [Fig. 29(b)]. It can be continued in this way until the desired number of channels is reached. The construction of the first channel, eliminating vertexes and EP edges from the surface, in other words, remains with a coordination lower than 12, which implies that there are free bonds. Type VP sites from the internal decahedron have two free bonds, sites VE have three and sites EP have four. So, surface sites are not only the difference between surface sites and the ones that

Table 20. Coordination or number of first neighbors (NN), of the different types of sites in the pentadecahedra with sites in shells in external, the same and internal layers.

Figure 27. Pentadecahedron of 2766 atoms of order 65, obtained from a decahedron without a central site of order 6 with 5 layers in the waist, which means adding 4 SEC-type layers to the original decahedron.

Crust

T

EE

EP

EV

RF

VE

VP

externa la misma interna

3 6 3

6 6 0

4 6 2

6 4 2

4 4 4

8 4 0

6 5 1

255

Crystallography and Shape of Nanoparticles and Clusters Table 21. Geometric characteristics for pentadecahedra with and without a central site, obtained from the corresponding decahedron of order . Without a central site NRF Nag N N

'

NAV

1

1 2

0 0

0 0

0 16

22 32

23 39

0 0

0 6

7 12

7 13

2

1 2 3

0 0 5

0 0 15

0 51 102

82 102 122

105 156 207

0 0 10

0 31 62

47 62 77

54 85 116

3

1 2 3

0 0 5

0 0 25

0 106 212

182 212 242

287 393 499

0 0 20

0 76 142

127 152 177

181 257 333

4

1 2 4

0 0 10

0 0 70

0 181 543

322 362 442

609 790 1152

0 0 60

0 141 423

247 282 352

428 569 851

1 2 3 5

0 0 5 15

0 0 45 135

0 276 552 1104

502 552 602 702

1111 1387 1663 2215

0 0 40 120

0 226 452 904

407 452 497 587

835 1061 1287 1739

5

NRF

With a central site Nag N N



Note: ' is the number of equatorial layers in the cluster. Number of sites AV , NAV , R sites NR , of sites added Nag , surface site N , and of total sites N in the cluster of order '. Notice that, for ' = 1, values of Table 1 are obtained. Even when ' can have any value higher than zero, here only some values are presented.

were eliminated, but the interior decahedron sites are added which have a coordination lower than 12. Table 22, lists the geometric characteristics for the decahedra, with and without a central site, modified with surface channels.  is the original cluster order. The only characteristics presented are for an n = 1, a single channel and 2, a triple channel. Eliminating the sites for the channel in the decahedron affects the number of sites SE and EC of the original decahedron. Table 22 lists the number of sites that are eliminated Nel , the number of sites SE and EC, per shell and in total, and the number of sites at the surface and in total in the final cluster. The number of sites SE and EC is needed in order to generate the decahedra with channels and developed, as was done for the pentadecahedron. From Table 22 are deduced the analytic expressions for the number of sites eliminated in a decahedron in order to obtain the decahedron with channels, the number of SE and EC sites, and the number of sites in the surface and the number of total sites in the cluster, which are presented next

(a)

Figure 29. Decahedron of 609 sites of order 4 and (a) a single channel, and (b) a triple channel per edge converging to the pole, of 532 sites and 412 sites, respectively.

for the decahedron with a central site and a simple channel: Nel  = 20 − 3 NtSE  = 5 − 5

N  =

N  = 20 2 − 18

20 3 44  + 10 2 −  + 4 3 3

for the decahedron with a central site and a triple channel: Nel  = 60 − 43

NSE  = 10 − 20

NtSE  = 5 2 − 15

NEC  = 10 − 15 N  = 20 2 − 58

NtEC  = 5 + 1 − 14 N  =

20 3 164  + 10 2 −  + 44 3 3

and for the decahedron without a central site and a simple channel: Nel  = 20 − 13 NtSE  = 5 − 1 − 4 N tEC  = 5 2 − 5 N  =

NSE  = 10 − 15 NEC  = 10 − 1 N  = 20 2 − 20 − 13

20 3 59  −  + 13 3 3

for the decahedron without a central site and a triple channel,

NtEC  = 5 2 − 15 N  =

Figure 28. (a) Decahedron of 287 sites, of order 3, with a central site,(b) Decahedron of 609 sites, of order 4, with a central site.

NEC  = 10 − 5

NtEC  = 5 + 1 − 4

NtSE  = 5 − 1 − 14

(b)

NSE  = 10 − 1

2

Nel  = 60 − 73

(a)

(b)

NSE  = 10 − 25 NEC  = 10 − 20 N  = 20 2 − 20 − 53

20 3 179  −  + 73& 3 3

In the same way used to generate the pentadecahedra, by modifying the decahedra, the new type of decahedra can be obtained, (Fig. 30). In order to do this, a layer of type SEC is added for each layer desired to increase the waist of the decahedron with channels. Table 22 lists the number of sites SE and EC, which are used to construct Table 23 for the decahedra with channels and developed.

256

Crystallography and Shape of Nanoparticles and Clusters Table 22. Geometric characteristics of decahedra with and without a central site, modified with surface channels. With a central site NtSE NEC NtEC

Without a central site



n

Nel

NSE

2

1

37

10

15

15

26

62

3

1 2

57 137

20 10

40 30

25 15

56 46

4

1 2

77 197

30 20

75 65

35 25

5

1 2

97 257

40 30

120 110

45 35

N

N

Nel

NSE

NtSE

NEC

NtEC

N

N

68

27

5

6

10

15

27

27

162 122

230 150

47 107

15 5

26 16

20 10

40 30

107 67

134 74

96 86

302 262

532 412

67 167

25 15

56 46

30 20

75 65

227 187

361 261

146 136

482 442

1014 854

87 227

35 25

96 86

40 30

120 110

387 347

748 608

Note:  is the cluster order. Here are presented only the characteristics for a single channel, n = 1, and a triple channel, n = 2. The number of eliminated sites Nel , the number of sites SE and EC per shell, and the number of sites in the surface and in total in the final cluster are listed.

When adding layers to widen the waist, the number of sites EE and VE is duplicated with respect to the original polyhedron, and different type EV and RF sites are generated, as happened with pentadecahedra, and the number of sites is given by

Nag  ' = ' − 1 10 − 1 with central site and a triple channel NRF  ' = 52 − 3' − 2

NEV  ' = 10' − 2 while the number of sites RF, the number of sites added Nag , of which Nag are added to the surface for ' equatorial layers for the polyhedron with central site and a single channel is: NRF  ' = 52 − 3' − 2 Nag  ' = ' − 110 2 + 5 − 9 Nag  ' = ' − 110 − 5 with a central site and a triple channel NRF  ' = 52 − 5' − 2 Nag  ' = ' − 110 2 + 5 − 39 Nag  ' = ' − 110 − 15 without central site and a single channel NRF  ' = 10 − 2' − 2

(a)

Nag  ' = ' − 110 2 − 5 − 9

(b)

Figure 30. Decahedron of order 4 of 609 sites with a central site, with (a) one single channel, and (b) with a triple channel per edge converging to the pole, with three layers in the waist, of 874 sites and 714 sites, respectively.

Nag  ' = ' − 110 2 − 5 − 39 Nag  ' = ' − 110 − 20&

3.4.2. Modified Decahedra with Surface Reconstruction Surface reconstruction of a decahedron is obtained if each triangular face of the decahedron proceeds as follows: a triangular face is obtained exactly as the one from the original decahedron; it is transported outwards to the same distance as the one between parallel triangular faces, 0&7953dNN , which is the height of a tetrahedron with a base of 1&05dNN from one side and the other sides of 1&0dNN , then in a parallel way is shifted to the waist to a distance of 0&606dNN , which is 2/3 of the height of the equilateral triangle of 1&05dNN from one side, until the generated sites coincide with the triangle centers of the original triangular face, and finally, the complete edge of the waist is eliminated. Hereafter these polyhedra are called mrdec (Montejano’s reconstructed decahedron) [33]. Figure 31 shows an example of a decahedron of 287 sites of order 3, to which 21 sites are added per face, this is 210 sites in total, for a mrdec of 497 sites. It should be noted that it seems that a decahedron with a surface channel is obtained, although the sites which seem to form the channel are at a distance of 1&13dNN , while in a decahedron with a simple channel they are at 1&63dNN . This is why, in this modification, a long bond is considered, and the EP sites of the interior decahedron do not form part of the surface. In fact, the surface of the resulting polyhedron is formed by the added sites and the sites VE, VP, and EE of the internal decahedron, which are those which also can be considered as surface because they remain with free bonds. It should be noted that the planes above the equator and the sites in the plane of the equator of the original decahedron are not modified. The characteristics of these polyhedra are presented in Table 24.  is the original cluster order,

257

Crystallography and Shape of Nanoparticles and Clusters

Table 23. Geometric characteristics for the decahedra with and without a central site, modified with surface channels and with added layers in the waist.



'

NEV

NRF

2

1 2 3 1 2 3 1 2 4 1 2 3 5

0 0 10 0 0 10 0 0 20 0 0 10 30

0 0 5 0 0 15 0 0 50 0 0 35 105

3

4

5

With a central site A single channel A triple channel Nag N N NRF Nag N 0 41 82 0 96 192 0 171 513 0 266 532 1064

62 77 92 162 187 212 302 337 407 482 527 572 662

68 109 150 230 326 422 532 703 1045 1014 1280 1546 2078

— — — 0 0 5 0 0 30 0 0 25 75

— — — 0 66 132 0 141 423 0 236 472 944

— — — 122 137 152 262 287 337 442 477 512 582

N

NRF

— — — 150 216 282 412 553 835 854 1090 1326 1798

0 0 10 0 0 20 0 0 60 0 0 40 120

Without a central site A single channel A triple channel Nag N N NRF Nag N 0 21 42 0 66 132 0 131 393 0 216 432 864

27 37 47 107 127 147 227 257 317 387 427 467 547

27 48 69 134 200 266 361 492 754 748 964 1180 1612

— — — 0 0 15 0 0 50 0 0 35 105

— — — 0 36 72 0 101 303 0 186 372 744

— — — 67 77 87 187 207 247 347 377 407 467

N — — — 74 110 146 261 362 564 608 794 980 1352

Note:  is the cluster order, ' is the number of equatorial layers added. Only the characteristics for a single channel, n = 1, and a triple channel, n = 2 are presented. Notice that for ' = 1 Table 22 values are obtained. Although ' can take any value higher than zero, here are only presented some values.

Nag is the number of sites added, N is the number of sites in the surface, and N is the total number of sites in the cluster. From Table 24, it is possible to obtain the analytic expressions as a function of the order of the original cluster for the different characteristics listed here, and they are presented next for the polyhedron with a central site: Nag  = 20 2 + 10 N  = 20 2 + 20 + 2 N  =

46 20 3  + 30 2 +  + 1 3 3

and without a central site

Alternatively, it is possible to generate the decahedra with surface reconstruction and developed. For this, simply add one SEC-type layer for each layer wanted to increase the waist of the decahedron with surface reconstruction, Figure 32. Table 19 presents the sites of type SE and EC of the original decahedron, used to construct Table 25 for the decahedra with surface reconstruction and developed. When adding the layers to widen the waist of the sites EE and VE of the original decahedron, they duplicate and generate the sites of type EV and RF, as occurred with the pentadecahedra, and the number of sites is the same. The number of added sites Nag from which Nag are added to the surface for ' equatorial layers for the polyhedron with a central site is Nag  ' = ' − 1 10 2 + 5 + 1

Nag  = 20 2 − 10

Nag  ' = ' − 110

N  = 20 2 − 3 N  =

29 20 3  + 20 2 − & 3 3

and without central site Nag  ' = ' − 1 10 2 − 5 + 1 Nag  ' = ' − 1 52 − 1&

Table 24. Geometric characteristics for the mrdec, decahedra with and without a central site, modified with surface reconstruction.

Figure 31. Decahedron of order 4 with 609 sites and with a central site, with 360 sites aggregated for a mrdec of 969 sites.



Nag

2 3 4 5

100 210 360 550

With a central site N N 122 242 402 602

205 497 969 1661

Nag 60 150 280 450

Without a central site N N 77 177 317 497

114 331 708 1285

Note:  is the cluster order. The number of added sites Nag , of surface sites N , and the total number of sites N in the cluster are listed.

258

Crystallography and Shape of Nanoparticles and Clusters

Figure 32. mrdec decahedron with a central site of order 4 (609 sites), with surface reconstruction (969 sites) and three equatorial layers (two layers SEC of 181 sites each one) for a total of 1331 sites in the polyhedron.

3.5. Truncated Decahedra (Marks Decahedra) These result from the adequate elimination of some sites of a certain decahedron. The resulting geometry is a figure of 22 vertexes (of three types), 40 edges (of 4 types, 15 from the original decahedron, but shorter, and 25 which are generated by elimination of the adequate sites), 10 pentagonal faces (triangular faces from the original decahedron are converted to irregular pentagons), and 10 equilateral triangular faces (at the equator and joint by pairs) (Fig. 33). The adequate elimination of sites is equivalent to eliminating the end sites of the edges which converge in the vertexes of the equator of the corresponding decahedron. Notice that in, each elimination, the equatorial edges lose two sites, while the edges which converge also toward the poles only lose one; this causes edges converging to the poles to be larger than the equatorial ones, but shorter then the ones from the original decahedron. Table 25. Geometric characteristics for the mrdec, decahedra with and without a central site, modified with surface reconstruction and added layers in the waist. With a central site Nag N N

Without a central site NRF Nag N N



'

NEV

NRF

2

1 2 3

0 0 5

0 0 15

0 51 102

122 142 162

205 256 307

0 0 10

0 31 62

77 92 107

114 145 176

3

1 2 3

0 0 5

0 0 25

0 106 212

242 272 302

497 603 709

0 0 20

0 76 152

177 202 227

331 407 483

4

1 2 4

0 0 10

0 0 70

0 181 543

402 442 522

969 1150 1512

0 0 60

0 141 423

317 352 422

708 849 1131

5

1 2 3 5

0 0 5 15

0 0 45 135

0 276 552 1104

602 652 702 802

1661 1937 2213 2765

0 0 40 120

0 226 452 904

497 542 587 677

1285 1511 1737 2189

Note:  is the cluster order, ' is the number of equatorial layers in the polyhedron. Notice that, for ' = 1, Table 24 values are obtained. Although ' can take any value higher than zero, here are presented only some values.

Figure 33. Truncated decahedron of 1372 sites of order 63, without a central site.

In the first step, n = 1, the equator vertexes from the equator of the exterior shell are eliminated, the last shell, 5 sites. In the second step, n = 2, two sites per equatorial edge are eliminated, and one of the rest of the edges, 20 sites of the previous stage are converted into 25. In the third step, n = 3, the step of the edges of the last shell is repeated, 20 sites, plus two sites of each triangular face, 20 sites, besides, in one shell before the last, the interior decahedron, the equator vertexes are eliminated, 5 more sites are needed in order to obtain 40 sites to be eliminated in this step, and there is a complete total of 70 eliminated sites in three steps. And so on, for n <  because NEE = 52 − 1 = 52 − 2 for decahedra with [without] a central site, for  = n − 1, there are only 3 [2] sites EE, see Table 26. Table 26 lists the truncated decahedra, resulting from decahedra with and without a central site, respectively. The order of the truncated decahedron consists of two numbers corresponding to the first and second columns of Table 26, the first column is the order of the decahedron generated , and the second column is the number of steps n needed to eliminate the adequate sites, or is half of the sites eliminated from each edge at the equator: n = 1 means that only the vertexes at the equator are eliminated, the first step in the elimination process. The third column lists the total number of sites eliminated from the original decahedron to obtain the truncated decahedron of order n, N− , in the fourth [eighth], the number of sites per equatorial site EE remaining in the originals are presented (taking into account that in these sites are included the two new vertexes VE) per edge, for the polyhedron with a central site [without a central site] (for n = 1 is the number of sites in edges in the original decahedron); finally, the fifth, sixth, and seventh columns [ninth, tenth, and eleventh] present the number of sites SE and EC and the total in the resulting truncated decahedron with a central site [without a central site]. For example, the truncated decahedron without a central site of order 63, Figure 33, is generated from the decahedron with a central site of order 6 (1442 sites, 10 sites per edge, 151 sites SE and 180 sites EC), and eliminating the 5 sites VE, 4 sites EE of each one of the equatorial edges of the surface, 2 sites EP from each edge toward the poles and the 5 VE sites from the immediate interior shell, so the truncated decahedron of order 63 has 6 sites per equatorial edge, 70

259

Crystallography and Shape of Nanoparticles and Clusters Table 26. Geometric characteristics for the truncated decahedra, constructed from a certain decahedra of order , with and without a central site.

 1

Order n 1

N− 5

With a central site N NEE NtSE NtEC 1

5

6

Without a central site N NEE NtSE NtEC

18









2

1

5

3

20

26

100

2

11

15

49

3

1 2

5 25

5 3

45 40

56 46

282 262

4 2

31 26

40 30

176 156

4

1 2 3

5 25 70

7 5 3

80 75 65

96 86 71

604 584 539

6 4 2

61 56 46

75 65 50

423 403 358

5

1 2 3 4

5 25 70 150

9 7 5 3

125 120 110 95

146 136 121 101

1106 1086 1041 961

8 6 4 2

101 96 86 71

120 110 95 75

830 810 765 685

6

1 2 3 4 5

5 25 70 150 275

11 9 7 5 3

180 175 165 150 130

206 196 181 161 136

1828 1808 1763 1683 1558

10 8 6 4 2

151 146 136 121 101

175 165 150 130 105

1437 1417 1372 1292 1167

Note: n is the half of sites eliminated from each equator edge.

sites are eliminated, and it has a total of 136 sites SE, 150 sites EC and, 1372 in total. The number of eliminated sites, the third column of Table 26, is the same for the two polyhedra, with and without a central site, and depends only on the number of steps n given and is obtained by 5 N− n = nn + 12n + 1& 6 Expressions for the number of sites SE, sites EC, and in total for the truncated decahedra with a central site 5 NtSE  n = 5 − nn − 1 2 2

5 NtEC  n = 5 + 1 + 1 − nn + 1 2 16 20 3  + 10 2 +  + 1 − N− n N  n = 3 3 and for the truncated decahedra without a central site 5 NtSE  n = 5 − 1 + 1 − nn − 1 2 5 NtEC  n = 5 2 − nn + 1 2 20 3   + − N− n& N  n = 3 3 The truncated decahedron is a polyhedron formed by 22 vertexes, joined by 22 vertexes, attached by 40 edges, forming 10 pentagonal faces and 10 triangular. Vertexes are of three types: VP, VE, and V ′ ; VP (2 vertexes), and are the same as in the original decahedron, VE and V ′ (10 vertexes each) result by pairs from elimination of the original VE and from elimination of sites from EE and EP; sites V ′ are

found to be the end of the edges that converge toward the poles, that is, the edges EP join sites VP and V ′ . Edges are of 4 types: the original EP edges (10 edges) and EE (5 edges), but the shorter ET (20 edges) and EV ′ (5 edges) which make up the triangular faces TF (10 faces) formed upon elimination of the equatorial sites, the edges EV ′ join the vertexes V ′ by pairs, and the edges ET join sites VE and V ′ . So, any truncated decahedron will have vertex sites of type VP, VE, V ′ , edges type EP, EE, ET, EV ′ , and faces type TF and pentagonal face PF. The number of VP sites is 2, of V′ is 10, and of VE is also 10. The number of remaining sites is variable, and is listed in Table 27 for the truncated decahedra with and without a central site. The number of surface sites is also listed as well as the total sites of the polyhedron. Columns 1 and 2 correspond to the order of the cluster  and n respectively. Columns 3–8 [11–16] correspond to sites EP, NEP , EE, NEE , ET, NET , EV ′ , NEV ′ , PF, NPF , and TF, NTF , respectively, and the two last columns to the surface sites N and the total of sites N for truncated decahedra with [without] a central site. From Table 27 it is possible to deduct the analytic expressions for the number of sites of the different types of sites as a function of the order of the truncated decahedron, the resulting expressions, common for the polyhedra with and without a central site, but depending on the number of steps made for the elimination of the sites, are NET , NEV ′ , and NTF NET n = 20n − 1 NEV ′ n = 5n − 1 NTF n = 5n − 2n − 1& The expressions for the other types of sites are as follows for the one with a central site: NEP  n = 102 − 1 − 10n NEE  n = 52 − 1 − 10n NPF  n = 102 − 1 − 1 − 10nn − 1 N  n = 20 2 − 5n2 + 2 and without a central site: NEP = 20 − 1 − 10n NEE = 10 − 1 − 10n NPF = 102 − 3 − 1 − 10nn − 1 N = 20 2 − 20 − 5n2 + 7&

3.5.1. Developed Truncated Decahedra The same method used to generate pentadecahedra is used here. Intermediate layers can be added to the truncated decahedra in order to obtain polyhedra with 15 lateral faces, 5 rectangular, and 10 trapezoidal (before triangular) (Fig. 34). This means that a layer of SEC type is added per each layer wanted to widen the polyhedron, so the polyhedron of Figure 34 has 4 equatorial layers; this means that 3 SEC

260

Crystallography and Shape of Nanoparticles and Clusters Table 27. Geometric characteristics for the truncated decahedra with and without a central site. The number of surface sites for each type of site, the number of total surface sites, and in the polyhedron are listed.



n

NET

NEV ′

NTF

NEP

With a central site NEE NPF N

N

NEP

Without a central site NEE NPF N

N

2

1

0

0

0

20

5

30

77

100

10

0

10

42

49

3

1 2

0 20

0 5

0 0

40 30

15 5

100 80

177 162

282 262

30 20

10 0

60 40

122 107

176 156

4

1 2 3

0 20 40

0 5 10

0 0 10

60 50 40

25 15 5

210 190 150

317 302 277

604 584 539

50 40 30

20 10 0

150 130 90

242 227 202

423 403 358

5

1 2 3 4

0 20 40 60

0 5 10 15

0 0 10 30

80 70 60 50

35 25 15 5

360 340 300 240

497 482 457 422

1106 1086 1041 961

70 60 50 40

30 20 10 0

280 260 220 160

402 387 362 327

830 810 765 685

layers were added. So, in a developed truncated decahedron, there will be the same sites as in one without development, except for the TF sites, triangular face, which convert into sites TR trapezoidal face, and the sites generated when adding the SEC layers, sites in the rectangular faces RF and in the vertical edges EV , joining the VE sites which are duplicated. The number of sites that will be added depends on the number of equatorial layers that the polyhedron will have, NtSE + NtEC sites from the original truncated decahedron are added per each layer to widen them. It is to be noticed that, in this case, the SE sites corresponding to EV ′ and ET sites will be added to the surface also, that is, they will not remain internal, along with the EC sites. Table 28 lists the number of sites of the different types of sites in the truncated decahedron with and without a central site and developed. When adding SEC layers to grow the truncated decahedron, the number of sites VP, EP, ET, V ′ , and PF are not modified, while the number of sites VE, and EE are duplicated, sites EV ′ increase and sites TF are converted to TR and they increase with respect to the original truncated decahedron. Sites RF and EV appear. From Table 28 it is possible to obtain the analytic expressions for the different number of sites in the developed truncated decahedron. The number of sites EV ′ , EV , and TR is common for the two

polyhedra with and without a central site, that is, they only depend on n and ', and their analytic expression is the following: NEV ′ ' n = 5n + ' − 2 NEV ' = 10' − 2 NTR ' n = 5n − 1n + 2' − 4 and the expressions for the other types of sites, for the one with a central site, NRF  ' n = 52 − 2n − 1' − 2 Nag  ' n = ' − 1 52 + 1 + 1 − 5n2  N  ' n = 102 + ' − 1 − 5n2 + 2 and without a central site NRF  ' n = 10 − n − 1' − 2 Nag  ' n = ' − 1 52 − 1 + 1 − 5n2  N  ' n = 102 + ' − 3 − 5n2 − 5' + 12& the total number of sites is equal to the total number of sites in the original decahedron plus the Nag sites.

3.6. Star-Type Decahedra

Figure 34. Truncated decahedron without a central site of order 63 with four equatorial layers, for a total of 2230 sites.

If the method to obtain a truncated decahedron is applied to a decahedron without a central site, which has an odd number of edges, it could be possible to eliminate all of the sites from the equatorial edges, except one, obtaining the star-type decahedron, named after the shape it represents (Fig. 35). For this decahedron, the equator edges from the original decahedron are practically eliminated, with one site remaining which converts to a vertex. The edges toward the poles are reduced to a half. The resulting figure has 17 vertexes, 30 edges, 10 rhombohedral-shaped faces, and 10 lateral triangular faces (over the equator and joint in pairs). The types of sites present are the same as in the truncated decahedron, although the PF sites, pentagonal faces, change to R sites, rhombohedral faces. The number of VP sites is 2, of V ′ is 10, and of VE is 5; the number of sites EP, ET,

261

Crystallography and Shape of Nanoparticles and Clusters Table 28. Geometric characteristics for the truncated decahedra with and without a central site and developed. With a central site Nag N N

Without a central site NRF Nag N N



n

'

NEV ′

NEV

NTR

NRF

2

1 1

1 2

0 5

0 0

0 0

0 0

0 46

77 97

100 146

0 0

0 26

42 57

49 65

3

1 1 1 2 2 2

1 2 3 1 2 3

0 5 10 5 10 15

0 0 10 0 0 10

0 0 0 0 10 20

0 0 15 0 0 5

0 101 202 0 86 172

177 207 237 162 192 222

282 383 484 162 248 334

0 0 10 0 0 0

0 71 142 0 56 112

122 147 172 107 132 157

176 247 318 156 212 268

4

1 1 1 1 2 2 2 2 3 3 3 3

1 2 3 4 1 2 3 4 1 2 3 4

0 5 10 15 5 10 15 20 10 15 20 25

0 0 10 20 0 0 10 20 0 0 10 20

0 0 0 0 0 10 20 30 10 30 50 70

0 0 25 50 0 0 15 30 0 0 5 10

0 176 352 528 0 161 322 483 0 136 272 408

317 357 397 437 302 342 382 422 277 317 357 397

604 780 956 1132 584 745 906 1067 539 675 811 947

0 0 20 40 0 0 10 20 0 0 0 0

0 136 272 408 0 121 242 363 0 96 192 288

242 277 312 347 227 262 297 332 202 237 272 307

423 559 695 831 403 524 645 766 358 454 550 646

Note: n is the cluster order, ' is the number of equatorial layers in the polyhedra. Notice that, for ' = 1, Table 27 values are obtained. Although ' can take any value higher than zero, here are only presented some values.

and EV ′ per edge is the same, not the number of edges of each type. Table 29 describes the geometric characteristics of the star-type decahedra, notice that they correspond to the truncated decahedron with a central site with ' =  (only one site remanins in the equatorial edges). From Table 29 it is possible to obtain the analytic expressions for the star-type decahedra, and they are presented next: NEP  = 10 − 1

NR  = 10 − 12

NET  = 20 − 1

NTF  = 5 − 2 − 1

NEV ′  = 5 − 1

N−  =

NtSE  =

5  + 1 2

N  = 15 2 + 2

To these decahedra, it is possible to add intermediate layers. The resulting figure has 10 trapezoidal lateral faces (which were triangular), and besides the 10 rhombohedral, 5 edges more, (Fig. 36). When adding layers of type SEC, the number of sites VP, EP, ET, V ′ , and RF, are not modified, while the number of sites EE and VE are duplicated; the number of sites EV ′ increases and the TF sites, triangular face, change to TR, trapezoidal face. Besides, sites EV appear. Table 30 lists the geometric characteristics of these polyhedra. Analytical expressions for the number of sites for the modified sites are presented next:

5  + 12 + 1 6 5 NtEC  =  + 1 + 1 2 9 15 N  = 5 3 +  2 +  + 1& 2 2

NEV ′  ' = 5 + ' − 2 NTR  ' = 5 − 1 + 2' − 4 NEV ' = 5' − 2 Nag  ' = ' − 1 5 + 1 + 1

Table 29. Geometric characteristics for the star-type decahedra.

Figure 35. Star-type decahedron of order 4.



NEP

NR

NET

NTF

NEV ′

N−

NtSE

NtEC

1 2 3 4 5 6 7 8 9

0 10 20 30 40 50 60 70 80

0 10 40 90 160 250 360 490 640

0 20 40 60 80 100 120 140 160

0 0 10 30 60 100 150 210 280

0 5 10 15 20 25 30 35 40

5 25 70 150 275 455 700 1020 1425

5 15 30 50 75 105 140 180 225

6 16 31 51 76 106 141 181 226

N

N

17 62 137 242 377 542 737 962 1217

18 80 217 459 836 1378 2115 3077 4294

262

Crystallography and Shape of Nanoparticles and Clusters

3.7. Additional Truncations in Decahedra

Figure 36. Star-type decahedron of order 4 and three equatorial layers added, with 661 sites.

 5    + 1  2 NtSE  = 5    + 1 + 1 2  5     + 1 + 1 2 NtEC  =    5 + 1 2

 impar  par  impar  par

N  ' = 5 − 13 + 2' + 1 + 10' + 7 N  ' = 5 3 +

Just as in the decahedra, it is possible to add or eliminate sites to obtain modified structures of all of the types of decahedra previously described to obtain new structures. This is achieved by eliminating the sites on the edges which converge to a pole in a truncated decahedron, for example, the one of order 74 without a central site with 2139 sites, a truncated decahedron with one channel per edge converging to a pole is obtained, Figure 37(a), of 2032 sites, having an elimination of 107 sites, sites EP, VP, EV , and V ′ . If only the sites EP, VP, and V ′ are eliminated, Figure 38(a), a polyhedron of 2047 sites is obtained. If, now, the following edges are eliminated, the result is a truncated decahedron with a triple channel of 1822 sites, 210 sites were eliminated [Fig. 37(b)], and also, it is possible to eliminate only the sites at edges without eliminating the EV sites [Fig. 38(b)]. This process can continue until obtaining the desired number of channels. Elimination of only the EP, VP, and V ′ sites does not affect the number of SE and EC sites from the original polyhedron, listed in Table 26, while upon elimination of the sites EV they do modify. Table 31 lists the geometric characteristics of some truncated decahedra with and without a central site and the resulting ones with single and triple channels when sites EP, VP, EV , and V ′ are eliminated. For the case when sites EP, VP, and V ′ are eliminated the geometric characteristics are listed in Table 32.

15 2 9  + 1 +  + 1 + Nag  ' 2 2

the total number of sites is equal to the total number of sites in the original star-type decahedron plus the Nag sites.

Table 30. Geometric characteristics for the star-type decahedra. 

'

NEV ′

NTR

NEV

Nag

NtSE

NtEC

N

N

1 2 2 3 3 3 4 4 4 4 5 5 5 5 5 6 6 6 6 6 6

1 1 2 1 2 3 1 2 3 4 1 2 3 4 5 1 2 3 4 5 6

0 5 10 10 15 20 15 20 25 30 20 25 30 35 40 25 30 35 40 45 50

0 0 10 10 30 50 30 60 90 120 60 100 140 180 220 100 150 200 250 300 350

0 0 0 0 0 5 0 0 5 10 0 0 5 10 15 0 0 5 10 15 20

0 0 31 0 61 122 0 101 202 303 0 151 302 453 604 0 211 422 633 844 1055

5 15 16 30 31 30 50 51 50 51 75 76 75 76 75 105 106 105 106 105 106

6 16 15 31 30 31 51 50 51 50 76 75 76 75 76 106 105 106 105 106 105

17 62 82 137 167 197 242 282 322 362 377 427 477 527 577 542 602 662 722 782 842

18 80 111 217 278 339 459 560 661 762 836 987 1138 1289 1440 1378 1589 1800 2011 2222 2433

(a)

(a′)

(b)

(b′)

Figure 37. Truncated decahedron without a central site of order 74 with (a) one single channel and 2032 sites, and (b) a triple channel and 1822 sites, eliminating AP , VP, AV , and V ′ sites. [(a′ ) and (b′ ) are lateral views of superior (a) and (b) views, respectively].

263

Crystallography and Shape of Nanoparticles and Clusters

for a triple channel, Nel  = 104 − 5 for a truncated decahedron without a central site and a simple channel, Nel  n = 20 − 5n − 13 and for a triple channel, (a)

Nel  = 104 − 7&

(a′)

And when only sites EP, VP, and V ′ are eliminated from a truncated decahedron with a central site and a simple channel, Nel  n = 20 − 10n + 2 for a triple channel, Nel  n = 104 − n − 3

(b)

for a truncated decahedron without a central site and a simple channel,

(b′)

Nel  n = 20 − 10n − 8

Figure 38. Truncated decahedron without central site of order 74 with (a) one single channel and 2047 sites, and (b) a triple channel and 1857 sites, eliminating AP , VP, and V ′ sites. [(a′ ) and (b′ ) are lateral views of superior (a) and (b) views, respectively).

and for a triple channel, Nel  n = 104 − n − 5 Also from Table 31 it is possible to obtain the analytical expressions for the total number of EC and SE sites when sites EP, VP, EV , and V ′ are eliminated from a truncated decahedron. It can be notice that the behavior of these quantities do not depend if the polyhedron has or not a central site. For the case of a simple channel,  5 n odd NtSE sc = NtSE Marks − 0 n even  0 n odd NtEC sc = NtEC Marks − 5 n even

In both Tables the number of eliminated sites and the total number of sites in the cluster are enumerated. Besides that in Table 31 the total number of sites SE and EC are also enumerated. From Tables 31 and 32, in both cases it is possible to obtain the analytical expressions for the total number of sites eliminated from these clusters. When sites EP, VP, EV and V’ are eliminated from a truncated decahedron with a central site and a simple channel Nel  n = 20 − 5n − 3

Table 31. Geometric characteristics for the truncated decahedra with and without a central site modified with a single and a triple channel, eliminating EP, VP, EV , and V ′ sites. With a central site Order  n 2 3 4

5

1 1 2 1 2 3 1 2 3 4

Nel 32 52 47 72 67 62 92 87 82 77

A single channel NtEC NtSE 15 40 40 75 75 60 120 120 105 95

26 56 41 96 81 71 146 131 121 96

N

Nel

68 230 215 532 517 477 1014 999 959 884

30 70 70 110 110 110 150 150 150 150

A triple channel NtEC NtSE 15 40 30 75 65 60 120 110 105 85

16 46 41 86 81 61 136 131 111 96

N

Nel

38 160 145 422 407 367 864 849 809 734

22 42 37 62 57 52 82 77 72 67

Without a central site A single channel A triple channel NtEC N Nel NtSE NtEC NtSE 6 26 26 56 56 41 96 96 81 71

15 40 25 75 60 50 120 105 95 70

27 134 119 361 346 306 748 733 693 618

10 50 50 90 90 90 130 130 130 130

6 26 16 56 46 41 96 86 81 61

5 30 25 65 60 40 110 105 85 70

N 17 84 69 271 256 216 618 603 563 488

264

Crystallography and Shape of Nanoparticles and Clusters

Table 32. Geometric characteristics for the truncated decahedra with and without a central site modified with a single and a triple channel, eliminating EP, VP, and V ′ sites. With a central site

Without a central site Channel

 2 3 4

5

Order n 1 1 2 1 2 3 1 2 3 4

Single Nel N 32 52 42 72 62 52 92 82 72 62

68 230 220 532 522 487 1014 1004 969 899

Triple Nel N 40 80 70 120 110 100 160 150 140 130

Single Nel N

28 150 150 412 412 387 854 854 829 769

22 42 32 62 52 42 82 72 62 52

Triple Nel N

27 134 124 361 351 316 748 738 703 633

20 60 50 100 90 80 140 130 120 110

7 74 74 261 261 236 608 708 583 523

Figure 40. Truncated decahedron with surface reconstruction.

N  n = and for a triple channel, 

0 10  10 NtEC tc = NtEC cs − 0 NtSE tc = NtSE cs −

n odd n even

and for the truncated decahedron without a central site Nag  n = 10 2 − 3 + 1 − ' − 1' + 2 + 1

n odd n even

The total number of sites in the cluster is the total number of sites in the truncated decahedron minus the eliminated sites. With this, it is possible to add the layers type SEC to obtain truncated decahedra with channels and developed [Figs. 39(a) and (b)]. Another modification of a truncated decahedron with surface reconstruction, as explained for the decahedron, can be achieved by shifting the pentagonal face outwards and to the waist, eliminating then the sites EE, VE, ET, and V ′ , which reflects in adding Nag sites (Fig. 40). Table 33 lists the geometric characteristics of the truncated decahedra modified by surface reconstruction. From this Table can be deduced that for the truncated decahedron with a central site the number of sites added, the number of surface sites and the total number of sites in the cluster are

20 3 46 5 25 65  + 30 2 +  − n3 − n2 − n + 1 3 3 3 2 6

N  n = 20 2 − 5n2 − 10n − 3 N  n =

29 5 25 65 20 3  + 20 2 −  − n3 − n2 − n& 3 3 3 2 6

When adding the sites for the surface reconstruction, the number of sites SE and EC is not affected. So, layers of type SEC can be added to obtain truncated decahedra with surface reconstruction and developed (Fig. 41). Table 34 lists the geometric characteristics for the truncated decahedron with and without a central site modified with surface reconstruction and with SEC layers added. From this table it is

Table 33. Geometric characteristics for the truncated decahedra with and without a central site modified with surface reconstruction.



n

With a central site Nag N N

Without a central site Nag N N

Nag  n = 10 2 2 +  − 2 − n − 1n + 2

3

2

150

202

412

90

137

246

N  n = 20 + 1 − 5nn + 2 + 2

4

2 3

300 240

362 327

884 779

220 160

277 242

623 518

5

2 3 4

490 430 350

562 527 482

1576 1471 1311

390 330 250

457 422 377

1200 1095 935

6

2 3 4 5

720 660 580 480

802 767 722 667

2528 2423 2263 2038

600 540 460 360

677 642 597 542

2017 1912 1752 1527

7

2 3 4 5 6

990 930 850 750 630

1082 1047 1002 947 882

3780 3675 3515 3290 2990

850 790 710 610 490

937 902 857 802 737

3114 3009 2849 2624 2324

(a)

(b)

Figure 39. Truncated decahedron without a central site of order 74 with (a) one single channel, and (b) a triple channel and three layers in the waist.

Note: n is the cluster order. Number of agreggated sites, Nag , of surface sites, N , and of the total of sites, N , in the polyhedron are listed.

265

Crystallography and Shape of Nanoparticles and Clusters

N  ' n =

16 5 20 3  +10 2 +  +1− nn+12n+1+Nag 3 3 6

and for polyhedra without a central site Nag  ' n = 10 2 −5 −10n+5'−1 N  ' n = 20 2 +5n2 −10'2 −10 −35'+10'+22 N  ' n =

1 5 20 3  ++  − nn+12n+1+Nag  ' n& 3 3 6

3.8. Icosahedra Figure 41. Truncated decahedron with surface reconstruction and with three SEC layers.

possible obtain expressions for the number of sites aggregated, the number of surface sites and the total number of sites in the cluster, for polyhedra with a central site they are Nag  ' n = 10 2 +5 −10n+6'−1 N  ' n = 20 2 +5n2 −10'2 +10 −20'+10'+2

Table 34. Geometric characteristics for the truncated decahedra with and without a central site modified with surface reconstruction and developed.



n '

3 3 3 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

2 2 2 2 2 2 2 3 3 3 3 2 2 2 2 2 3 3 3 3 3 4 4 4 4 4

1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

With a central site Nag N N 0 91 182 0 166 332 498 0 156 312 468 0 261 522 783 1044 0 251 502 753 1004 0 241 482 723 964

233 234 235 393 404 415 426 418 429 440 451 593 614 635 656 677 618 639 660 681 702 653 674 695 716 737

262 353 444 584 750 916 1082 539 695 851 1007 1086 1347 1608 1869 2130 1041 1292 1543 1794 2045 961 1202 1443 1684 1925

Without a central site Nag N N 0 60 120 0 125 250 375 0 115 230 345 0 210 420 630 840 0 200 400 600 800 0 190 380 570 760

177 142 87 317 292 247 182 342 317 272 207 497 482 447 392 317 522 507 472 417 342 557 542 507 452 377

156 216 276 403 528 653 778 358 473 588 703 810 1020 1230 1440 1650 765 965 1165 1365 1565 685 875 1065 1255 1445

Note:  is the cluster order. Number of agreggated sites, Nag , of surface sites, N , and of the toal of sites, N , in the polyhedron are listed.

The icosahedron (ICO) is a geometric figure with a central site, 12 vertexes, connected with 30 edges, and formed by 20 triangular faces; Figure 42 shows a 561 icosahedron. From this figure, it can be seen that the icosahedron can also be considered as two pentagonal pyramids with a rotation of 36 of one with respect to the other, and with intermediate layers, to adequately join the vertexes of both pyramids. Each icosahedral cluster presents a certain number of sites, but distributed in different layers, which are formed by equivalent sites: sites at the same distance from the origin, which have the same environment, and the same type of neighbors. Sites in the ICO are localized in triangular faces (T ), edges (E), and vertexes (V ). The order-1 ICO is formed by one central site and one first shell with only one layer of 12 vertexes. The second-order ICO is formed by the addition of one shell formed by 42 sites, to complete 55 sites, distributed in two layers: one layer of 30 sites E and a second layer of 12 sites V . When adding a third shell of 92 sites, the ICO of third order of 147 sites is completed, the added sites are distributed in three layers: one of 20 sites T , one more of 60 sites E and the third of 12 sites V . And successively, complete shells are added, forming clusters of order . The geometric characteristics of the ICO are listed in Table 35, to clusters of order 10. The first column lists the order of the cluster . The following columns list the number of sites in each type of sites in the cluster. Next, is the number of sites of each site or the layers of each type of site. Finally, the number of sites in the shell and in the total

Figure 42. Icosahedron of order 5 with 561 atoms.

266

Crystallography and Shape of Nanoparticles and Clusters

Table 35. Geometric characteristics for icosahedra. 

NT

NE

NV

RT

RE

RV

 Nshell

T Nshell

1 2 3 4 5 6 7 8 9 10

0 0 20 60 120 200 300 420 560 720

0 30 60 90 120 150 180 210 240 270

12 12 12 12 12 12 12 12 12 12

0 0 1 1 2 3 4 5 7 8

0 1 1 2 2 3 3 4 4 5

1 1 1 1 1 1 1 1 1 1

1 2 3 4 5 7 8 10 12 14

1 3 6 10 15 22 30 40 52 66

N 12 42 92 162 252 362 492 642 812 1002

N 13 55 147 309 561 923 1415 2057 2869 3871

of sites in the cluster is given. From this table, it is possible to obtain the dependence on the cluster order  from the total number of sites N , on the number of surface sites N , the number of sites T NT and of sites E NE , as well as the number of layers of sites T RT , and sites E RE , which are expressed as 10 3 11  + 5 2 +  + 1 3 3 NT  = 10 − 1 − 2

N  = 10 2 + 2

N  =

Rm T  =

m/2 m/2−1      3"# 3" + a +    "=1 "=1   

NE  = 30 − 1

    m−1/2    1+a   + 6" + a#  1 + a "=1  /2# RE  =  − 1/2#

m even   = 3m + a# a = −1 0 1# m odd

 even  odd&

3.9. Decmon-Type Polyhedra In contrast with previous polyhedra, decmon polyhedra are based on a decmon-type pyramid, which is constituted by 16 vertexes joined by 30 edges which form 15 faces: 5 rectangular type (100) and 10 triangular of type (111), 5 of which are equilateral and 5 isosceles; Figure 43. The 5 equilateral triangular faces (TE) have m edge sites per side, and each

side of length 1&05m + 1dNN , where dNN is the distance to the first neighbors; the 10 isosceles triangle faces (TI) have n edge sites per side, two sides of length n + 1dNN and one of length 1&05n + 1dNN ; the 10 rectangular faces (RF), with m × n sites, have sides 1&0n + 1dNN and 1&05m + 1dNN . Equal sides of faces TI coincide with the sides of faces RF, and the different side forms part of the equator along with one of the sides of RF faces. Faces TE coincide with the pole, and have a common side; also, each face has a common face with one RF face. The angle between the planes of RF and TE faces is 190&66 , and between RF and TI faces is 161&94 . The 16 vertexes (V ) are of three types: 1 pole, where the 5 TE faces converge, 5 where the three types of faces (V 3) converge, and 10 at the equator, and where an R and a TI face converge (V 4). These vertexes are joined by 30 edges (E) of 5 types: 5 of length 1&05m − 1, between TE faces and which converge at the poles and join the VP and V 3 vertexes (EE); 5 of length 1&05m − 1, between TE and RF faces and join vertexes V 3 (E3); 10 of length 1&0n − 1 between TI and RF faces and join vertexes V 3 and V 4 (E34); and 10 at the equator, 5 of them of length 1&05m − 1 of one RF (ER) face and 5 of length 1&05n − 1 of one TI (EI) face, which joins V 4 vertexes. Based on the decmon pyramid, many polyhedra can be obtained: the truncated icosahedron, the decmon, and the decmon with intermediate aggregated layers.

3.9.1. Decmon Icosahedra The truncated icosahedron or icosahedron of decmon type is a polyhedron obtained upon adequate truncation and in a symmetric way of an icosahedron (however, they can be truncated asymmetrically also). An asymmetric truncation of an icosahedron is done in the following manner: surface sites forming a cover with a pentagonal pyramid shape over a vertex, yielding a decmon-type pyramid, with rectangular faces and a side with a length of dNN . In order to perform a symmetrical truncation, first a cover like that previously presented has to be eliminated, but from the vertex diagonally opposite. The following truncations have to be done by elimination of sites which form a pyramid-shaped cover of decmon type. Figure 44 shows two icosahedra with their corresponding truncations. Figure 44(a) [Fig. 44(d)] shows a 561 (309) site icosahedron that, upon truncation, yields the 409 (207) sites T Ih (observe in this one the decmon-type pyramid); Figure 44b [Fig. 44(e)], which, when truncated, again yields the 257 (105) sites T Ih , Figure 44(c) [Fig. 44(f)]. From this figure, it can be observed that it is possible to obtain the T Ih of two different kinds: (1) the T Iha [Fig. 44(f) and (b)], and (2)the T Ihb [Figs. 44(b), (c), and (e)].

3.9.2. Truncated Icosahedra TIha

Figure 43. Decmon type pyramid.

As shown in Figure 44(f), it is formed by 22 vertexes, joined by 50 edges, forming faces of three types: rectangular, RF type (110), and two triangular of type (111), one equilateral (TE) and the other isosceles (TI). Also, from this figure it can be noticed that the T Iha can be formed by attaching two decmon-type pyramids, with TI faces coinciding with TI faces at the equator or waist.

267

Crystallography and Shape of Nanoparticles and Clusters Table 36. Geometric characteristics for truncated icosahedra T Ih . 

(a)

(d)

(b)

(e)

Figure 44. (a) Icosahedron Ih of 561 atoms. (b) Truncated icosahedron, T Ih of 409 atoms. (c) T Ih of 257 atoms. (d) Ih of 309 atoms. (e) T Ih of 207 atoms. (f) T Ih of 105 atoms.

3.9.3. Truncated Icosahedra T Ihb As shown in Figures 44(b), (c), and (e), it is formed by 32 vertexes, joinrd by 80 edges of I order to form 50 faces. 10 vertexes, 30 edges, and 20 faces more than T Iha . The 10 vertexes form one plane, and are of equator type from T Iha and the 20 edges connect the vertexes from the two planes forming the 20 extra faces of trapezoidal type TR, now having the waist of the T Ihb . T Ih are characterized by three numbers m n r where m n r is the number of sites in one edge between faces TE, (RF and TI) [TR], including vertexes, for example, in Figure 44(f), there is T Iha 3 3 1, and in Figures 44(b), (e), and (f), there are the T Ihb 5 2 4, 4 3 2, and 4 2 3. It is clear that, based on this notation, to the icosahedron corresponds an m = r and an n = 1, for example, 6 1 6 and 5 1 5 from Figures 44(a) and (d), respectively. From Figure 44, it can be seen also that T Iha results from complete truncation (when there is only one or two edge sites from the original icosahedron) of an icosahedron with an odd number of sites in the edges. By truncation of an icosahedron, only some T Ih are obtained, and upon truncation of another icosahedron, others are obtained; in each truncation, a certain number of sites are eliminated, but the same number in each truncation. Table 36 presents some of the geometric characteristics of T Ih . The first column shows the order of the icosahedron from where the  comes, the second column shows the number of truncations done nt , the third column shows the three characteristic numbers, in the fourth column is the number of sites eliminated from the original icosahedron n− , in the fifth column the number of sites remaining in the surface N is shown, and in the sixth column, the total number of sites in T Ih , N is shown. So, for example, the T Ih (5, 3, 3), comes from the icosahedron of order 6, to which 424 sites were eliminated in two truncations, yielding a T Ih of 499 sites, of which 242 are at the surface.

mnr

n−

N

N

1

0

2,1,2

0

12

13

2

0 1

3,1,3 2,2,1

0 32

42 22

55 23

3

0 1

4,1,4 3,2,2

0 62

92 62

147 85

4

0 1 2

5,1,5 4,2,3 3,3,1

0 102 204

162 122 82

309 207 105

5

0 1 2

6,1,6 5,2,4 4,3,2

0 152 304

252 202 152

561 409 257

6

0 1 2 3

7,1,7 6,2,5 5,3,3 4,4,1

0 212 424 636

362 302 242 182

923 711 499 287

(c)

(f)

nt

Note: nt is the number of truncations made.

From Table 36 it can be seen that the maximum number of truncations is /2  − 1/2, for  even [odd]; besides the values for m, n, r, and m + n + r show the next characteristics  + 2/2  even ≤m≤+1  + 3/2  odd   + 2/2  even 1≤n≤  + 1/2  odd 1  even ≤r ≤ 2  odd  + 3  even ≤ m + n + r ≤ 2 + 1 + 1&  + 4  odd It is worth also notice that the parity of r is equal to the one for , and the maximum values of m and r coincide with the minimum value of n; m and n take values of one to one, while r takes values of 2 to 2. The analytical expressions for n− , N and N are

5 + 1 n−  = 2nt 1 + 2

N  = N ico − 10nt = 10 2 − 10nt + 2 N  = N ico − n−

11 5 + 1 10 3 2  + 5 +  + 1 − 2nt 1 + & = 3 3 2 For the truncated icosahedra T Iha , r = 1 and m = n =  + 2/2,  even, based in the structure itself there are sites on vertexes VP, V 3, and V 4, on edges EE, E3, E34, ER, and EI, on triangular faces TE and TI, and on squared faces RF. The number of sites on vertexes are 2, 10 and 10, respectively. The number of sites for the other type of sites,

268

Crystallography and Shape of Nanoparticles and Clusters

which is given in function of m, and the number of surface atoms can be obtained from the following expressions NEE = NE3 = 10m − 2

NER = NEI = 5m − 2

NE34 = 20m − 2

NTE = NTI = 5m − 2m − 3

NR = 10m − 22

(a)

N = 20m − 12 + 2&

For the truncated icosahedra T Ihb , the number of sites V 4, ER, and EI are duplicated with respect to the original polihedron, the number of surface sites is modified and the sites EL and TR, sites on lateral edges and on trapezoidal faces, are aggregated, and they are given by NEL = 10r − 2  0 r =2 NTR = 10 nr − 1 − n − 1 r ≥ 3   5m + n − 2m + n − 1 + 2 r = 2 N = 5m + n − 2m + n − 1   + 10r + nr − 2n − 28 r ≥ 3&

3.9.4. Decmon Polyhedra

A decmon polyhedra, Figure 45, is very similar to T Iha , it has the same number of vertexes, edges, and faces, and the same type of sites. Also, it can be constructed by two decmon-type pyramids. The difference is that, at the equator or at the waist, faces TI have TI faces, and RF have RF faces. The lengths of decmon edges, even when they are of three sizes, are related to the values m and n of the decmon pyramid, characterizing each particular decmon. mn can be considered as the order of the decmon (decmon mn), and the values taken by m and n are not limited. m and n are the (number−1) of edge sites EE and E34, respectively. Indeed, every decmon is formed by centered shells in one decmon of an order in which: (1) m = 1 and n = 1, (2) m > 1 and n = 1, or (3) m = 1 and n > 1, which are the smallest. For example, decmon of Figure 45 is of order 78, decm78, which in turn is a decm67 + one shell, decm67 is a decm56 + one shell, decm56 is a decm45 + one shell, decm45 is a decm34 + one shell, and decm34 is decm23 + one shell, decm23 is a decm12 + one shell, decm12 being the smallest of this family. Decmon of order 11 (decm11) is the smallest of all decmons, and the edges at the equator are all the same;

(b)

(c)

Figure 46. Decmon-type decahedron of order 11. In (b) the sizes of the atoms of (a) were reduced to more clearly observe the squared and triangular faces by bondings to first neighbors, and to show that this is a polyhedron with a central site. (c) Decmon-type decahedron of order 44, decm44, of 609 sites.

Figure 46. It only has sites in vertexes, and is centered in one site, having a cluster of 23 sites. There is a decmon family mm, based on decm11, and the possible decmons to be created are decm22, decm33, decm44, decm55, and so on. The geometric characteristics of a decahedra-type decmon of order mm are listed in Table 37. The first column lists the m value, the four following columns list the number of sites in each type of sites, rectangular face sites RF, in triangular faces T , in edges E, and in vertexes V . It should be clear that T sites include TE and TI sites, which in number are equal. Besides, the 5 types of edges are included in E, corresponding 1/5 to each one of EE and E3, 1/10 to ER, and another to EI and 2/5 to E34. Vertexes are always 22, and from the three kinds already mentioned, 2 poles, 10 V 3, and 10 V 4. The following three groups of two columns list the number of sites in the plane above the equator, at the equator, and for the number of sites in clusters, for each shell and in total. Figure 46(c) shows the decmon for m = 4 and of 609 atoms. From Table 37, it can be seen that, for a decmon of order m, the dependency with m values of the number of sites in a rectangular face NRF , triangular face NT , edges NE , surface N , and total N is given by the following expressions: NRF = 10m − 12

NT = 10m − 1m − 2 N = 20m2 + 2

NE = 50m − 1

N =

20 3 16 m + 10m2 + m + 1 3 3

Table 37. Geometric characteristics for decmons of family of order mm based on decmons of order 11, decm11.

Figure 45. Decmon-type polyhedron.

m

NRF

NT

NE

NV

NSE

NtSE

NEC

NtEC

1 2 3 4 5 6 7 8 9 10

0 10 40 90 160 250 360 490 640 810

0 0 20 60 120 200 300 420 560 720

0 50 100 150 200 250 300 350 400 450

22 22 22 22 22 22 22 22 22 22

5 10 20 30 40 50 60 70 80 90

5 15 35 65 105 155 215 285 365 455

10 20 30 40 50 60 70 80 90 100

11 31 61 101 151 211 281 361 451 551

N

N

22 82 182 322 502 722 982 1282 1622 2002

23 105 287 609 1111 1833 2815 4097 5719 7721

269

Crystallography and Shape of Nanoparticles and Clusters

also, expressions for the number of sites SE and EC can be calculated for each m, NSE , NEC , and total NtSE and NtEC : NSE = 10m − 1

NtSE = 5m2 − m + 1

NEC = 10m

NtEC = 5mm + 1 + 1&

The next family decmon size, following the decmon family mm, is the decmon family mn, with n = m − 1 or n = m + 1, and based on the decm12 and decm21, respectively. Decm12 and decm21 have the same number of sites, a total of 54, although decm21 is more elongated toward the poles; Figure 47. Decm12 has 20 sites in edges E34 and 5 in EI, while decm21 has 10 sites in edges EE, 10 in E3, and 5 in ER, and both are centered in a decahedron of 7 sites, regular for decm12 and more elongated toward the poles for decm21. From decm12, other decm can be generated: decm23, decm34, decm45, decm56, and so on. From decm21, the decm generated are decm32, decm43 decm54, decm65, and so on. The geometric characteristics of decmon-type decahedra, of order mn, with m = n − 1 and m = n + 1, are listed in Table 38. The first two columns list the m and n values, and in the following 10 columns, the number of sites of each type of site, which considers only the sites in rectangular faces RF, equilateral triangular faces TE, isosceles triangular faces TI, of two types of edges: short edges Em, and large edges En, and in vertex V . It should be clear that sites in Em [En] include sites EE, ER, and E3 [EI and E34], corresponding (2/5) to EE, (2/5) to E3, and (1/5) to ER [(4/5) to EI and (1/5) to E34]. The vertexes are always 22, and come from the three types already mentioned. The number of sites RF and V is common for the two decmon mn families, with m = n − 1 and m = n + 1, while the number of sites TE and TI, Em, and En are interchanged. Columns 8–11 correspond to the interchange of values of m and n; this is what happens with decmon families of order mn with m = n + 1. The following three groups of two columns list the number of sites in the plane above the equator (SE sites), at the equator (EC sites), and the numbers of sites in the cluster, per shell and in total, these values are common to the m and n values and are interchanged. Figures 47(c) and (d) show decmons of order 45 and 54, of 835 atoms. After the mn decmon families, with m = n − 1 and m = n + 1, follow in size the ones of decmon of order mn, with m = n − 2 and m = n + 2, and are based on decm13 and decm31, respectively. The decm13 and decm31 decmon have the same number of sites, 105 in total, although decm31 is more elongated toward the poles, Figure 48. Decm13 has 40 sites in edges E34, 10 in EI, and 10 in faces TI, while decm31 presents 20 sites in edges EE, 20 in E3, 10 in ER,

and 10 in faces TE, and both are centered at a decahedron of 23 sites with central site, regular for the decm13 and more elongated toward the poles for the decm31. From the decm13, it is possible to generate the decm24, decm35, decm46, decm57, and so on. From decm 31, it is possible to obtain the decm42, decm53, decm64, decm75, and so on. The geometric characteristics of the decahedra type decmon of order mn, with m = n − 2 and m = n + 2, are listed in Table 39, with the same indications for the columns as in Table 38. Figures 48(c) and (d) show the 1111 atoms decmon of order 46 and 64. From Tables 37–39, it is deduced that, for a decmon order mn, the dependency on the m and n values of the number of sites in rectangular face NRF , sites in the equilateral triangular face NTE , of sites in an isosceles triangular face NTI , of sites in the edge type Em (EE, E3, and ER), NEm , of sites in the edge type En (EI and E34) NEn , and of sites in the surface N is given by the following expressions: NRF = 10m − 1n − 1

NTE = 5m − 1m − 2

NTI = 5n − 1n − 2

NEm = 25m − 1 N = 5m + n2 + 2&

NEn = 25n − 1

Also, it is possible to obtain the number of sites in the immediate superior plane over the equator for each m and n, NSE and in total NtSE , and in the equator NEC and NtEC for the number of sites in the surface and the number of sites for each m and n and in total for the decmon for each combination of m and n used here. For m = n ± 1 NSE = 10m

NtSE = 5mm+1+1

NEC = 10m+5

NtEC = 5m+12  20 3 61   m +20m2 + m+7#   3 3 for + N=      20m3 + 1m# for − 3 3

N = 20m2 ±20m+7

and for m = n ± 2 NSE = 10m+5

NtSE = 5mm+2+5

NtEC = 5m2 +15m+11  20 136   m3 +30m2 + m+23#   3 3    for + N = 20m2 ±40m+22 N =  20 16   m3 −10m2 + m−24#    3 3 for −

NEC = 10m+1

3.9.5. Modified Decmon (a)

(b)

(c)

(d)

Figure 47. Decmon-type decahedrons of (a) order 12, and (b) order 21 with 54 sites. Decmon-type decahedron of 835 sites of (c) order 45, decm45, and (d) order 54, decm54.

To obtain decmon type decahedra, widened with extra layers, the same procedure is followed as with the decahedra, separating the decmon as two pyramids with parallel bases, making the vertexes coincide, and adding layers of type SEC, Figure 49. The obtained figure has 10 regular lateral faces besides the ones previously mentioned, and 10 more edges.

270

Crystallography and Shape of Nanoparticles and Clusters

Table 38. Geometric characteristics for decmons of family of order mn with m = n − 1 and m = n + 1 based on decmons of order 12, decm12, and of order 21, decm21.

m 1 2 3 4 5 6 7 8 9 10

n

NRF

NTE

NTI

mn NEm

NEn

NTE

NTI

nm NEm

NEn

NV

NSE

NtSE

NEC

NtEC

2 3 4 5 6 7 8 9 10 11

0 20 60 120 200 300 420 560 720 900

0 0 10 30 60 100 150 210 280 360

0 10 30 60 100 150 210 280 360 450

0 25 50 75 100 125 150 175 200 225

25 50 75 100 125 150 175 200 225 250

0 10 30 60 100 150 210 280 360 450

0 0 10 30 60 100 150 210 280 360

25 50 75 100 125 150 175 200 225 250

0 25 50 75 100 125 150 175 200 225

22 22 22 22 22 22 22 22 22 22

10 20 30 40 50 60 70 80 90 100

11 31 61 101 151 211 281 361 451 551

15 25 35 45 55 65 75 85 95 105

20 45 80 125 180 245 320 405 500 605

Besides to the type of sites for decmon-type polyhedra, the modified decmon presents the lateral and vertical edges sites EV , sites on rectangular faces that join triangular faces RI, and sites on rectangular faces that join rectangular faces RR. The number of sites V 4, ER, and EI are duplicated with respect to the original polyhedron, and the number of surface sites is modified. The analytical expressions for them are the following NV 4 = 20

NEm = 30m−1

NEn = 30n−1

NEV = 10'−2

NRI = 5n−1'−2

NRR = 5m−1'−2

N = 5m+nm+n+1+2+5'−2m+n& The number of sites that are aggregated to the surface and to the total number of sites in the cluster for n = m ± 1 are given for Nag = 10m2 + 15m + 6' − 1 Nag = 10m + 5' − 2 and for n = m ± 2 Nag = 10m2 + 25m + 16' − 1 Nag = 10m + 1&

4. SYNTHESIS OF NANOPARTICLES Metal clusters have been formed in a supersonic beam, this method to prepare colloidal metals in nonaqueous media was first reported by the group of Andres at Purdue

(a)

(b)

(c)

(d)

Figure 48. Decmon-type decahedra of (a) order 13 and (b) order 31. Decmon-type decahedron of 1111 sites of (c) order 46, decm46, and (d) order 64, decm64.

N

N

47 127 247 407 607 847 1127 1447 1807 2207

54 181 428 835 1442 2289 3416 4863 6670 8877

University [34–36]. A supersonic beam source was used; this instrument produces metal clusters with a controlled mean diameter in the range of 1–20 nm in size. Once the clusters have nucleated, the cluster aerosol is directed through a spray chamber to be placed in contact with a fine spray of organic solvent and surfactant. However, conceptually, the simplest method of preparing colloidal metals is the condensation of atomic metal vapor into a dispersing medium [37]. Given the high oxidation potential of most atomic metals (for example, the oxidation potential for atomic gold is −1&5 V), the use of water as the diluent’s phase can be ruled out, so exclusively inert organic liquids are going to be used in this procedure. Since the activation energy for the agglomeration of metal atoms is very low, the possibility for competing molecular complex formation processes which have higher activation energies can be mitigated by operating at low temperatures. The use of metal vapors cocondensed with organic vapors to prepare colloidal metals in nonaqueous media was first reported by Roginski and Schalnikoff in 1927 [38], some 50 years before the recent wave of activity in metal vapor chemistry. The organosols were prepared at reduced pressure by the evaporation of relatively volatile metals such as cadmium, lead, and thallium, and a subsequent cocondensation of these with the vapors of organic diluents such as benzene and toluene on a liquid air cooled cold finger. After cocondensation was complete, a colloidal suspension of the metal was obtained by warming up the frozen matrix and collecting the liquid (Fig. 50). A new metal vapor synthesis system for the preparative scale cocondensation of metal vapors with aerosols of organic liquids was recently achieved by Bradley’s group [39]. The use of aerosol overcomes one limitation of the other methods, in which the organic diluent must be either volatile or has a liquid range which extends to a temperature low enough to reduce its vapor pressure to a useful value. The aerosol droplets, ca. 1'm in diameter, are generated by feeding the organic liquid (neat liquid, polymer solutions, solutions of nonvolatile ligands) into an ultrasonic atomizing nozzle from which they fall onto a rotating plate cooled to 77 K in a vacuum chamber, as shown in Figure 50. Vapors of one or more metals, obtained by simultaneous sputtering from metal or alloy targets, are then cocondensed with the aerosol. This results in the formation of a frozen organometallic matrix, which is then warmed up to allow the

271

Crystallography and Shape of Nanoparticles and Clusters

Table 39. Geometric characteristics for decmons of family of order mn with m = n − 2 and m = n + 2 based on decmons of order 13, decm13, and of order 31, decm31.

m 1 2 3 4 5 6 7 8 9 10

n

NRF

NTE

NTI

mn NEm

NEn

NTE

NTI

nm NEm

NEn

NV

NSE

NtSE

NEC

NtEC

3 4 5 6 7 8 9 10 11 12

0 30 80 150 240 350 480 630 800 990

0 0 10 30 60 100 150 210 280 360

10 30 60 100 150 210 280 360 450 550

0 25 50 75 100 125 150 175 200 225

50 75 100 125 150 175 200 225 250 275

10 30 60 100 150 210 280 360 450 550

0 0 10 30 60 100 150 210 280 360

50 75 100 125 150 175 200 225 250 275

0 25 50 75 100 125 150 175 200 225

22 22 22 22 22 22 22 22 22 22

15 25 35 45 55 65 75 85 95 105

20 45 80 125 180 245 320 405 500 605

20 30 40 50 60 70 80 90 100 110

31 61 101 151 211 281 361 451 551 661

aggregation of the metal atoms. The resulting colloid solution is removed from the reactor under helium atmosphere for characterization and further studies. In the 1980’s, several research groups in Japan prepared metal particles of Al, Mg, Mn, Be, Te, Fe, Pb, Co, Ni, Cd, Ag, In, Pd, and Au by this method [40–49]. The work most recently performed was by Uyeda in 1991 [50]. This study demonstrated the feasibility of producing films of almost any metal or semiconductor by metal vapor deposition in a noble gas high-pressure atmosphere. Another method to produce nanoparticles is the bioreduction method. Several researchers established that live and nonlive biologic systems such as algae have the ability to absorb metal ions from solutions through their cell walls and various cellular constituents [51, 52]. Studies have shown that algae are able to bind gold ions from aqueous solutions and form colloidal particles on their surfaces, yet the mechanisms have not been fully understood [53–57]. Several plants have been studied for their unique biochemical ability to accumulate metal compounds over extended periods. In fact, plant species such as Douglas-fir and rye grass are utilized as biological indicators of geologic gold deposits [58, 59]. Some other plants such as Indian mustard have been utilized for phytomining; the accumulation of valuable metals comes from low concentrations in soils by plants [60, 61]. Furthermore, Lujan and co-woekers reported that a purple color, similar to “Purple of Cassius”, resulted when aqueous

N 82 182 322 502 722 982 1282 1622 2002 2422

N 105 287 609 1111 1833 2815 4097 5719 7721 10143

Au(III) was reacted with the biomaterials in their study, indicating the formation of gold colloids [62–64]. Plants may possess a unique natural chemical stabilization mechanism that allows the formation of nanoparticles. In addition, the plant biomaterials may be isolated for a more thorough study of the colloidal information mechanisms. Therefore, by taking advantage of the naturally occurring compounds found within the plant systems, a novel method may be found to generate similarly sized stable metal colloids. While consistency in nanoparticle size and shape is important to many materials, differences in nanoparticle synthesis may also lead to changes in particle conformation and spatial arrangement, which may provide better nanoscopic building blocks to form new raw materials with distinctly different properties than their macroscaled counterparts. Recently, it was shown that live plants can uptake metal solutions and produce nanoparticles [65]. This approach is very promising for further developments in nanotechnology. All of the particles reposited in the present work were produced by the methods described above. A third way to prepare metal nanoparticles is the colloidal methods discussed in the introduction. In the original Brust reaction [13], the addition of dodecanethiol to the organic-phase AuCl− 4 (1:1 molecule), followed by reduction Liquid Reservoir

Feed Pump Piezoelectric atomizing nozzle

Metal vapor flux monitor and feedback Sputtering Gun

Vacuum chamber

Frozen Matrix Liquid Nitrogen

Rotating cold stage Liquid Nitrogen reservoir

Metal sol Liquid Nitrogen Ferrofluid seal

Figure 49. Decmon-type decahedron of order 78 with 6 intermediate layers added.

Figure 50. Schematic of sputtering source metal vapor/aerosol for metal colloid preparation with permission from [39], J. S. Bradley, 1994. © 1994, VCH, Weinheim.

272

Crystallography and Shape of Nanoparticles and Clusters

with BH− 4 , led to dodecanethiolate protected Au clusters having a 1–3 nm range of core diameters − AuCl− 4 toluene + RSH → −AuSR n polymer

−AuSR− 4 + BH4 → Aux SRy Subsequent reports have shown that a wide range of alkanethiolate chain lengths (C3–C24) [66], 1-functionalized alkanethiolates, and dialkyl disulfides [67] can be employed in this same protocol. An example of a passivated nanoparticles formula is Au145 (S(CH2 )5 CH3 )50 . Pt, Rh and Pd salts can be used following the same method. One of the possible reactions is − AuCl− 4 toluene + PtCl + R − SH

→ −Aux Pty SR− n polymer Aux Pty SR− 4 + BH4 → Aux Pty SRz The details of these reactions have not been completely understood; however, the behavior of the reaction is consistent with a nucleation-growth-passivation process: (1) larger thiol: gold mole ratios give smaller average nanoparticle metal core sizes [68, 69], and (2) fast reductant addition and cooled solutions produce smaller, more monodisperse nanoparticles [70, 71], and (3) quenching the reaction immediately following reduction produces higher abundances of very small core sizes (≤ 2 nm). The Schiffrin reaction tolerates considerable modification with regard to the protecting ligand structures. While alkanethiolate passivated nanoparticles (a) are nonpolar, highly polar ligands (b–d) can yield water-soluble passivated nanoparticles in modified synthesis. Nanoparticles passivated with arenethiolate [72–75] (e–g) and (2–mercaptopropyl) trimethyloxysilane ligands (h) have also been prepared. Sterically bulky ligands tend to produce smaller Au core sizes (relative to alkanethiolate-passivated nanoparticles prepared using equal thiol/AuCl− 4 ratios), suggesting a steric connection (as yet unproven) with the dynamics of core passivation [76]. The core metal of passivated nanoparticles can also be modified. Alloy nanoparticles with Au/Ag, Au/Cu, Au/Ag/Cu, Au/Pt, Au/Pd, and Au/Ag/Cu/Pd cores [77] have been reported. It is well known for the case of bare particles that noncrystallographic structures are formed in many cases, these being predominantly icosahedral and decahedral. In other cases fcc, bcc, or single twinned particles are also produced. Extensive literature is found on shape observations and the kinetic growth conditions in which the different types of particles are produced [78–90].

5. TRANSMISSION ELECTRON MICROSCOPY (TEM) AND SIMULATIONS OF IMAGES OF NANOPARTICLES 5.1. Introduction Since its inception in the 20th century, electron microscopy has developed into a powerful tool for scientific research. Indeed, the use of electron microscopy has facilitated many

fundamental discoveries. This technique has itself been subject to constant evolution thanks to unceasing design efforts in the scientific and manufacturing transmission electron microscopy (TEM) community. The advances in this field have been characterized by a series of quantum leaps in technology. For instance, the double condenser lens and tilting stages revolutionized the study of defects in metals in the late 1950s and early 1960s. The high lattice resolution achieved in the 1970s, along with improvements in vacuum methods and lens control, produced a less complicated instrument capable of achieving rapid results with minimal training. Consequently, the TEM (Transmission Electron Microscopy) gained enormous popularity in the materials community. In the second half of the 1980s, a new quantum leap was realized by the achievement of atomic resolution on a routine basis. Microscopes with a point resolution of 0.17 nm coupled with spectacular advances in diffraction theory led to many fine examples of materials characterization. In the 1990s another quantum leap occurred with the introduction of highly coherent field emission electron sources and electron loss analyzers. Combined with X-ray analysis, this established the TEM as a powerful analytical machine. For the first time, images, diffraction, and spectroscopy could be obtained from a single instrument. The TEM has become a standard instrument for many scientific fields. During the second half of the 1990s, the substantial increase in the number of papers published in Physical Review Letters containing electron microscopy studies attests to this fact. Other important developments include the introduction of electron holography and dark field images produced using incoherently scattered electrons (ZContrast). The next quantum leap in electron microscopy will be produced by the introduction of aberration correctors for the objective lens and STEM lens-forming system, and by the development of monochromators for the incident beam. These techniques will undoubtedly usher in a new era for TEM. Improved point resolution of 0.007 nm, an increased information limit, enhanced energy resolution, and improved stages using MEMS technology will allow 3D reconstruction of amorphous materials such as glasses. Refined X-ray techniques will facilitate chemical analysis essentially at the single atom level. Still more exciting, s corrected microscopes will allow more space in the objective lens gap, thus making it possible to construct stages for the in-situ examination of materials at atomic resolution and analysis with atomic accuracy [91]. The most recent developments in TEM coincide with the emergence of nanotechnology [92], and because the TEM probe size is ideal for nanoscale studies, it is clear that advanced TEM will be a major instrument in subsequent nanotechnological developments. The TEM will also be fundamental in areas of rapid development such as advanced materials research, biotechnology, and microelectronics. The automation of modern TEM has led to applications in semiconductor fabrication processes, and energy filtering will permit the study of polymers and biological materials with unprecedented accuracy. Transmission electron microscopy is one of the most important techniques to study nanoparticles. The resolution of the TEM is well below the size of most of the

273

Crystallography and Shape of Nanoparticles and Clusters

particles and in the order of magnitude of the metal–metal bonds. There are several papers which have explained in a very detailed way the electron microscopy methods to study nanoparticles. We will not discuss in this chapter dynamical diffraction methods, but we rather refer the reader to those publications [93–96]. Images were obtained using a JEOL 2010-FEG microscope, with a resolution of 0.19 nm and a tilting possibility of ±30 , and a JEOL 4000EX with a resolution of 0.17 nm. HREM (High Resolution electron microscopy) images of nanoparticles were obtained at the first optimum defocus (Scherzer focus) or at the second optimum defocus condition, resulting in images of atomic columns as black (first maximum) or white (second maximum) dots. Molecular Dynamics software was used to build theoretical models of the particles. Simulated images were obtained using the Simula TEM software of the Institute of Physics at UNAM, which is especially adequate for studying nanoparticles with noncrystallographic symmetries. A comparison of experimental and simulated data will allow us to interpret the experimental images. In general, in order to recognize a nanoparticle in a nonambiguous fashion, it is necessary to obtain a tilting sequence of HREM. Figure 51 illustrates the changes in contrast when the orientation of the particle is changed with respect to the electron beam. This case corresponds to a cubooctahedral particle. This image illustrates a remarkable difference between nanoparticles and large crystals. In the latter case, the image will be very sensitive to tilting, and will disappear with small tilts. The nanoparticle, on the other hand, always shows images despite the large angle of tilting. In many cases, the images correspond to pseudolattice images. Considerable caution should be taken when analyzing fringe images of nanoparticles. The images out of a lowindex direction do not necessarily reflect a structural feature. In diffraction terms, the finite size of the nanoparticles in real space results in real rods in reciprocal space. The Ewald sphere will always cut the reciprocal space points despite the angle. β j

[001]

5.2. fcc Particles The fcc particles are straightforward to identify. Figure 52 illustrates the contrast produced by an fcc by a pyramid formed by two tetrahedrons and their diffraction pattern (which in most instances is identical to the FFT, Fast Fourier Transform, of the image). Experimental examples of a pyramidal and of a tetrahedral particle are shown in Figure 53.

5.3. Decahedral Particles Very important cases to consider are the images produced by pentagonal particles such as the one shown in Figure 54. Typical TEM images of pentagonal particles are shown in Figure 55. The fivefold symmetry of the FFT can be observed. However, these patterns are a combination of the diffraction patterns of different portions of the particle, as shown in Figure 56. This indicates that the overall structure of the particle has a fivefold symmetry but this is not the result of fivefold symmetry in each portion of the particle. Because of that, it is more correct to consider that the particle has a pseudofivefold symmetry. This is in sharp contrast with the case of quasicrystals [97]. It should be remembered that another important effect in the images observed in a TEM is due to the defocus condition which can change the image significantly. Figure 57 illustrates this effect for the case of a regular decahedron. In this case the calculations were made for a JEOL 4000 microscope. As is known from image theories [98], there are two defocus conditions in which the image represents the true atomic columns which correspond on the figure to −40&5 nm (black atoms) and −70&2 nm (white atoms). If another defocus is used, the image does not necessarily reflect the atomic positions, but useful information still can be obtained. For instance, a defocus of −35&5 increases the grain boundary contrast, and the nature of the interfaces can be examined. At an even lower defocus of −20&5 nm, the pentagonal particle structure is reflected on the image. An experimental example of a pentagonal particle showing well-defined atomic resolution is shown in Figure 58. In this case, some of the twin boundaries are incoherent. An example of the structure formed at a defocus condition below the optimum value is shown in Figure 59.

90

β 75

α i

60

45

30

15

0

15

30

45

60

75

90 α

Figure 51. Simulation of HREM images of a cubooctahedral particle in different orientations with respect to the electron beam.

Figure 52. Models of a fcc pyramid formed by two tetrahedra. Observe their contrast in the second row and their diffraction pattern in the third row.

274

Crystallography and Shape of Nanoparticles and Clusters

b

a

10 nm

c

d 2 nm

Figure 55. High-resolution images showing atomic contrast of ∼2 nm pentagonal gold nanoparticles and their corresponding FFT showing the fivefold symmetry.

a

b

Figure 53. Bright field images of: (a) pyramidal particle, (b) Tetragonal particle, (c) FFT of (b), (d) diffraction pattern of (b).

A commonly observed type of pentagonal particle corresponding to a rounded shape and the respective model of the particle are shown in Figures 60(a) and (b), respectively. A complete set of images at different tilting angles is shown in Figure 61. Obtaining different images at different tilting angles and referring to this sequence should be enough data to fully identify the structure of that particle. In the previous sections, we described a new type of pentagonal structure that was described originally by Montejano [99] and was termed a decmon type of particle. This type of structure belongs to a more general family of decahedral nanoparticles, which includes the truncated decahedron

2 nm

Figure 56. (a) Pentagonal particle showing the FFT of different sections. (b) Overall FFT.

-605 Å

10 nm

Figure 54. Bright field image of a pentagonal gold particle.

-355 Å

-555 Å

-305 Å

-505 Å

-255 Å

-455 Å

-405 Å

-205 Å

Figure 57. Effect of defocus on TEM images. Calculations made for a truncated decahedron and for a JEOL 4000 microscope.

275

Crystallography and Shape of Nanoparticles and Clusters

a

b

2 nm 50.00 Angstroms

Figure 60. Pentagonal particle with rounded shape. (a) HREM image, (b) model of particle.

j

2.5 nm

β α

Figure 58. HREM image of a pentagonal particle, showing incoherent twin boundaries indicated by arrows.

i k

(Marks) [100], the pancake structure of Koga and Sugarawa [101], and the truncated icosahedron of Ascencio et al. [102]. A tilting sequence for three different decmon structures is shown in Figures 62–64. We also include the calculation a 64-order pentadecahedron in Figure 65. An experimental example of this type of particle is shown in Figure 66. This image corresponds to a dark field taken in a weak beam condition. In this case, a continuous set of thickness fringes is expected. However, because of the decmon shape, a pentagonal profile is formed at the center of the particle. Another experimental example is shown in Figure 67(a). The particle marked by an arrow can be identified using a model with surface reconstruction, as shown in Figure 67(b). We have found that surface reconstruction particles are formed very frequently when the sample approaches the equilibrium condition. For instance, the particle in Figure 67(a) was obtained after heating the sample with an electron beam and allowing recrystallization. In this process, the total surface energy of the nanoparticle will tend to be a minimum. There are several ways to achieve this surface energy minimization. Marks described truncations on the decahedron structure [100] that result in a more stable structure.

β 50.00 Angstroms

10

α

Figure 61. Images of a pentagonal particle obtained as a function of the tilting angles " and 4. The origin of the coordinate system indicates a 0, 0 position, and the range of the angles is 90 . The model and typical bright field and HREM image (including FFT) are shown in the left portion of the figure.

β 90 j β

75 α i 60

k 50.00 Angstroms

a

b 45

50.00 Angstroms

30

DECM10_3 15

0 2 nm

Figure 59. (a) High-resolution image of a pentagonal particle with a defocus below the optimal value. (b) Corresponding FFT indicating the fivefold symmetry.

α 0

15

30

45

60

75

90

Figure 62. HREM images calculated for a decmon 10 structure shown in the left portion of the figure. The rotation angles " and 4 go from 0 to 90 .

276

Crystallography and Shape of Nanoparticles and Clusters β β

90 j β

α α 75

i

i k

k

60

50.00 Angstroms

β 45

30

DEC1215

0

α 15

0

45

30

75

60

90

α

Figure 63. HREM images calculated for a decmon 12 structure shown in the left portion of the figure. The rotation angles " and 4 go from 0 to 90 .

Figure 65. HREM images calculated for a pentadecahedral structure as the model shown in the left portion of the figure. The rotation angles " and 4 go from 0 to 90 . a

However, the energy landscape has many structures with similar energy. Another possibility is to introduce faceting along the twin boundaries of the decahedra, resulting in the structure shown in Figure 68(a). If images are formed at the Scherzer condition, the channels can be visualized [Fig. 68(b)], in contrast to the case of the regular decahedron. We have described these particles previously in this chapter. A full map of images for these channeled structures as a function of the tilting angle for the optimum defocus is shown in Figures 69(a) and (b). This type of decahedron is often observed in larger particles, as shown in Figure 70. Another very interesting possibility has been suggested by molecular dynamic simulations of Chushak and Bartell [103] and other group [104]. This work indicates that, in the twin boundaries, a stacking fault is formed. In other words, the

j

b

5 nm

2 nm

Figure 66. (a) A weak multibeam dark field image of a decmon type of decahedron obtained in a gold particle synthesized by bioreduction. (b) An HREM image of a decmon-type particle.

a

90 β α 75 i 60

k 80.00 Angstroms

2 nm

45 b

30

15

0 0

15

30

45

60

75

90

Figure 64. HREM images calculated for a decmon structure with less degree of truncation, as the model shown in the left portion of the figure. The rotation angles " and 4 go from 0 to 90 .

30.00 Angstroms

Figure 67. Analysis of a decmon type particle supported on a thin portion of a carbon film (indicated by an arrow). (a) TEM Image, (b) Identification of particle using a model with surface reconstruction.

277

Crystallography and Shape of Nanoparticles and Clusters

a

cerius2 HRTEM simulation e=400 . 0kv Cs=1.00mm Ap=0.70A-1 Drms=250.0A div=0.3mrad vib=0.35A As=0.0A at 0.0

b 70 60 50 40 30

1 nm

10 nm

1 nm

20 10 0 0

50.00 Angstroms

20

40

60

t=50 . 00A def=-405 . 5A

Figure 68. (a) Model of decahedron showing faceting along the twin boundaries. (b) Simulated TEM image of a decahedron in Scherzer condition, where the channels can be observed.

ABC stacking will be converted to ABAB stacking in the boundary region. A different situation corresponds to the structure reported in Figure 68. In this case, the stacking is altered only in the last layer of the packing. This is illustrated in β 90 j

β α

75

i

60

k

45

30

Figure 70. (a) Bright field image of a large particle produced by bioreduction showing a typical contrast of channels in the twin boundaries in the interface. (b) and (c) show the high-resolution image of decahedral shape in which the contrast at the boundaries does not change with defocus and can be explained by the model of interface channels discussed in the theoretical section.

Figure 71 in which a regular stacking in a unit of the decahedra is altered in the surface. In other words, the ABCA sequence becomes ABCB. In Figure 71, each layer is illustrated in a different color. The yellow atoms correspond to the surface sites. We have found experimentally that this type of decahedron is very stable, especially at small sizes. The contrast of this particle is very peculiar, and is shown in Figure 72 for a different defocus. In this case, the boundaries show a very strong contrast at all defocus conditions. This provides a practical way to distinguish this particle from other types of decahedra which, under some focus conditions, can produce boundary contrast, as shown before. The contrast of the boundaries will be very strong, as shown in Figure 72. This type of particle is often observed experimentally in nanoparticles with sizes ∼1&5 nm, as shown in Figure 73(a), or when two metals are present. Figures 73(b)

a

MA74S3_

b

15

0 0

15

30

45

60

75

90

α

β 90 j

10.00 Angstoms

β

75 α

6:DD5C

10.00 Angstoms

c

i 60 k 45 30

15

0 α 0

15

30

45

60

75

90

Figure 69. (a) Map of images for the truncated decahedron as function of the tilting angle for the optimum defocus.

Figure 71. (a) Packing of atoms in a decahedron showing a stacking fault in the last atomic layer. Different colors correspond to different layers of the packing sequence, and the ABCA sequence becomes ABCB. (b) Another view of the same packing, in which the surface atoms are drawn in a smaller size. (c) Side view of the resulting stricture in which a missing atom in the center can be observed.

278

Crystallography and Shape of Nanoparticles and Clusters –705

–455

–505

–555

–605

–655

–205

–255

–305

–355

–405

1.5 nm

Figure 72. Sequence of simulated images of a decahedron particle with a missing atom in the center (fivefold site) as the result of surface recontruction. The image is calculated at a different defocus.

Figure 74. Distorted decahedral particle.

and (c) show the HREM image of a particle of Au-Pd at 50%–50% in which, in many places, it is possible to observe the surface atom vacancy. In general, it can be said that the central atom vacancy is a unique feature of this type of particle. Finally in some cases irregular or asymmetric decahedral shape is produced. This was described by Uppenbrink and Wales [105] and most likely is the result of the kinetics of growth; a typical example is shown in Figure 74. When the particles grow larger some special shapes are seen as indicated in Figure 75.

that, under flat surfaces, particles will be tilted until one triangular face (111) becomes in contact with the substrate. This results in most cases in a twofold orientation. However when the particles are around 1–2 nm, it is possible to observe the fivefold orientation more frequently. This is probably due to truncations on the surface in contact with the substrate. An example of this contrast and an experimental example are shown in Figure 79. When a bimetallic particle is synthesized, the frequency of the icosahedral particle increases, particularly at the smallest sizes [107].

5.4. Icosahedral Particles

5.5. Single Twinned and Complex Particles

One of the most frequently observed particles at small sizes corresponds to the icosahedral structure. A typical TEM image of these particles is shown in Figure 76. The profile is typically hexagonal in the threefold orientation. The most common images are in the two- and threefold orientation. The calculations of these images are shown in Figure 77 and are consistent with previous calculations [106]. Experimental images of a gold icosahedral particle in a threefold orientation and a twofold orientation are presented in Figures 78(a), (b) and (c), (d) including the FFT, which indicates the symmetry. A less commonly observed orientation is the fivefold. This is probably due to the fact

Twinning is a very general phenomenon which might happen in the very early stages of growth and does not necessarily result in a decahedral or icosahedral particle. In many cases,

a

b

1 nm

c

2 nm

2 nm 10 nm

Figure 73. Decahedral particles, in which a defect was formed in the last layer of atoms. Notice that the particle in (a) does not have a central atom in the fivefold site. The particles in (b) and (c) correspond to an Au-Pd system, and atom vacancies in several sites are observed.

Figure 75. Large irregular decahedral particle grown from a distorted decahedral particle.

279

Crystallography and Shape of Nanoparticles and Clusters

a

b

1 nm

c

d

10 nm

Figure 76. Bright field of an icosahedral gold particle showing a typical hexagonal profile. 5 nm

we can observe a pyramidal particle with a single twin. This particle is shown in Figure 80. We have observed particles down to 1.5 nm in size which present single twins. Therefore, it is safe to assume that, in some cases, twinning proceeds in the very early stages of growth. However, when the particles grow to larger sizes, coalescence can occur, leading to a more complex type of particles. We have found by molecular dynamics that a double icosahedral structure can be produced very often. In this structure, we have two interpenetrating icosahedra to form a double icosahedron, as shown in Figure 81. This structure

Figure 78. Experimental images of icosahedral particles and their corresponding FFT. (a), (b) threefold orientation. (c), (d) twofold orientation.

20.00 Angstroms

a

b

2 nm

Figure 79. Model and image simulations of an icosahedral particle in fivefold symmetry (upper portion) and experimental images for a gold particle in the lower part. In both cases, we include the FFT.

c

d

2.5 nm

Figure 77. Calculated images of HREM icosahedral particles at (a) threefold orientation, and (b)–(d) different twofold orientations.

Figure 80. HREM image of a single twin particle and its corresponding FFT.

280 a

Crystallography and Shape of Nanoparticles and Clusters

LIST OF SYMBOLS

b



Figure 81. (a) Model for a double icosahedron obtained by molecular dynamics. (b) simulated image. As can be seen, one part of the structure is a fivefold orientation and the other one is at fivefold orientation.

z N N  Nshell T Nshell S T H RF R TR PF TF TI TE E EE EV ET EI ER EP ES

2 nm

EH Figure 82. HREM image of a gold particle showing surface faceting.

corresponds to a minimum in the total energy, and has been observed experimentally by Hofmeister’s group [108].

5.6. Surface Faceting in Nanoparticles In many cases, the particles can present a surface faceting, especially when they are annealed or heated in-situ. This faceting is relevant since it can be related to the catalytic activity of the particles; an example is shown in Figure 82. The presence of steps will increase the particle curvature, yielding an almost spherical particle. We have observed that faceting results in (110) and (112) faces.

6. CONCLUSIONS We have made a systematic description of the most significant geometries which are formed in nanoparticles. We have calculated the number of different atomic sites and growth characteristics of the nanoparticles. It has been shown that the models correspond closely with the experimental HREM and diffraction data. New types of particles have been described in this work, which correspond to a new family of decahedral structures. We expect that, as the field advances and new synthesis routes for particles ∼1 nm are developed, all the if predicted clusters will be observed. This work also illustrates the richness of the crystallography of nanoparticles and clusters. Unlike the bulk materials, fivefold symmetry in nanoparticles produces a very rich landscape of structures.

EV ′ E34 V V′ VE VP V3 V4 SE EC SEC NI NtSE NtEC fcc fccc OCTAC OCTTC fccs OCTAS OCTTS bcc DODE CO ICO T Ih decmon

Order of the cluster of Number of crusts in the cluster Total coordination Number of surface sites Total number of sites in the cluster Number of layers in the shell  Total number of shells in the cluster of order  Squared face Triangular face Hexagonal face Rectangular face Rhombohedral face Trapezoidal face Pentagonal face Equatorial triangular face Isosceles triangular face Equilateral triangular face Edge Edge on the equator Vertical edges on the equator Edges between PF and TF Edge between TI faces Edge between R faces Edge toward pole sites Edges between square and hexagonal faces in truncated octahedrons Edges between hexagonal faces in truncated octahedrons Edges joining V’ Edges joining V3 and V4 Vertex Vertex at the end of truncated EP Vertex on the equator Vertex on pole sites Vertex where converge TI, TE and RF Vertex where converge TI and RF Sites in a plane just over the equator Sites on the equator Layer of the total of SE sites plus the total of EC sites Number of I sites, I= S, T, H, R, RH, Total number of SE sites in the cluster Total number of EC sites in the cluster Face-centered cubic Face-centered cubic with a central site Octahedron with a central site type fccs Truncated octahedron with a central site type fccs Face-centered cubic without a central site Octahedron without a central site type fccs Truncated octahedron without a central site type fccs Face-centered cubic Dodecahedron cubooctahedron Icosahedron Truncated icosahedron Decahedron of Montejano, described originally by J. M. Montejano-Carrizales

Crystallography and Shape of Nanoparticles and Clusters

mrdec TEM STEM HREM MEMS FFT

Montejano’s reconstructed decahedron, described originally by J. M. Montejano-Carrizales and J. L. Rodr´ıguez-L´opez Transmission electron microscope/microscopy Scanning transmission electron microscope/microscopy High resolution electron microscope/microscopy Micro-electro-mechanical system Fast Fourier transform

GLOSSARY Equivalent sites Sites located at the same distance from the origin, which occupy the same geometric place and have the same environment, that is, the same number and type of neighbors. EC layer The layer formed by the total of sites in the plane of the equator of decahedral structures. Geometric characteristics The set of the number of the different type of sites, the number of sites for each type of site, the number of sites by shell, of surface sites and the total of sites for each structure. SE layer The layer formed by the total of sites in the plane just above the equator of decahedral structures. SEC layer The layer formed by the joint between the SE layer and the EC layer. Site coordination The number of first nearest neighbors of the site with layers of its same shell, with interior and exterior shells, and it will be characteristic of each site. Standard coordinates Site coordinates described by triads (a b c), a, b and c are integers, where the total number of sites is obtained by doing permutations and commutations of the positive and negative values of a, b and c.

ACKNOWLEDGMENTS The authors are indebted to Mr. Luis Rendón and to Mr. J. P. Zhou for support with electron microscopy, to Mr. Samuél Tehuacanero for image processing. We acknowledge to CONACyT for full support to J.L.R.L., and partial support for J.M.M.C. through grants G-25851-E and W-8001 (Millennium Initiative). To Drs. Patricia Santiago, Jorge Ascencio, and Margarita Marín for useful discussions, to José Luis Elechiguerra for helpful modeling and simulations, and Victor Kusuma for help in preparing the manuscript. We also thank the office of V.P. of Research of the University of Texas at Austin for financial support.

REFERENCES 1. M. Gross, “The Nanoworld, Miniature Machinery in Nature and Technology,” p. 254. Plenum Trade, New York, 1999. 2. H. S. Nalwa, “Handbook of Nanostructured Materials and Nanotechnology,” Vol. 1–4, Academic, San Diego, 2000. 3. S. Sugano and H. Koizumi, “Microcluster Physics,” SpringerVerlag, Berlin, 1998. 4. S. a. P. Das-Sarma, “Perspectives in Quantum Hall Effects,” p. 430. Wiley, New York, 1997. 5. H. L. Stormer, Solid State Comun. 107, 617 (1998).

281 6. H. Grabert, “Single Charge Tunneling,” Plenum, New York, 1992, p. 335. 7. M. H. Dvoret, D. Esteve, and C. Urbina, Nature 547 (1992). 8. R. C. Ashoori, Nature 413 (1996). 9. A. P. Alivisatos, Science 933 (1996). 10. A. Rosen, Nanostructured Materials Made from Self-Assembled Particles, Chalmers University of Sweeden, 2000. 11. H. Haberland, “Clusters of Atoms and Molecules I and II: Solvation and Chemistry of Free Clusters, Embeded, Supported and Compressed Cluster,” Springer-Verlag, Berlin, New York, 1994, p. 412. 12. W. A. de Heer, Rev. Mod. Phys. 65, 611 (1993). 13. M. Brust, M. Walker, D. Bethel, D. Schiffrin, and R. Whyman, J. Chem. Soc. 801 (1994). 14. A. C. Templeton, M. J. Hostetler, E. K. Warmoth, S. Chen, C. M. Hartshorn, V. M. Krishnamurty, M. D. E. Forbes, and R. W. Murray, J. Am. Chem. Soc. 120, 4845 (1998). 15. R. Whetten, T. Khoury, M. Alvarez, S. Murthy, I. Vezmar, Z. L. Wang, P. W. Stephens, C. L. Cleveland, W. D. Luedtke, and U. Landman, Adv. Mater. 8, 428 (1996). 16. T. G. Schaaff, M. N. Shafigullin, J. T. Khoury, I. Vezmar, R. L. Whetten, W. Cullen, P. N. First, C. Gutierrez-Wing, J. Ascencio, and M. José-Yacamán, J. Phys. Chem. B 101, 7885 (1997). 17. D. V. Leff, L. Brandt, and J. Heat, Langmuir 12, 4723 (1996). 18. S. L. Logunov, T. S. Ahmadi, M. A. El-Sayed, J. T. Khoury, and R. L. Whetten, J. Phys. Chem. 101, 3713 (1997). 19. C. J. Kiely, J. Fink, M. Brust, D. Bethell, and D. J. Schiffrin, Nature 396, 444 (1998). 20. A. Mews, A. V. Kadavanich, U. Banin, and A. P. Alivisatos, Phys. Rev. B 53, 13242 (1996). 21. H. Bonnemann, W. Brijoux, R. Brinkmann, R. Fretzen, T. Joussen, R. Koppler, B. Korall, P. Neiteler, and J. Richter, J. Molec. Catal. 86, 129 (1994). 22. Z. L. Wang, Adv. Mater. 10, 13 (1998). 23. B. Korgel and D. Fitzmaurice, Phys. Rev. B 59, 14191 (1999). 24. H. Bonnemann and R. M. Richards, Eur. J. Inorg. Chem. 2455 (2001). 25. S. Ino and S. Ogawa, J. Phys. Soc. Jpn. 22, 1365 (1967). 26. S. Ogawa, S. Ino, T. Katar, and H. Ota, J. Phys. Soc. Jpn. 21, 1963 (1966). 27. T. P. Martin, Phys. Rep. 273, 199 (1996). 28. J. M. Montejano-Carrizales and F. Aguilera-Granja, Rev. Mex. Fís. 44 S3, 59 (1998). 29. J. M. Montejano-Carrizales, F. Aguilera-Granja, and J. L. MoránLópez, Nanostructured Materials 8, 269 (1997). 30. J. M. Montejano-Carrizales, M. P. Iñiguez, and J. A. Alonso, J. Cluster Sci. 5, 287 (1994). 31. L. García González and J. M. Montejano-Carrizales, Phys. Stat. Sol. A 230, 357 (2000). 32. M. Foiles, M. I. Baskes, and M. S. Daw, Phys. Rev. B 33, 1285 (1983). 33. J. M. Montejano-Carrizales and J. L. Rodríguez-López, Institute of Physics University of San Luis Potosi, Internal report (2002) and personal communication. Surface reconstruction and decahedral structure of bimetallic nanoparticles, S. L. Rodríguez-López, J. M. Montejano-Carrizales, U. Pal, J. F. Sánchez-Ramírez, D. García, M. Miki-Yoshida, and M. José-Yacamán, to be published. 34. R. Andres, J. Bielefeld, J. Henderson, D. Janes, V. Kolagunta, C. Kubiak, W. Mahoney, and R. Osifchin, Science 273, 1690 (1996). 35. D. Janes, V. Kolagunta, R. Osifchin, J. Bielefeld, R. Andres, J. Henderson, and C. Kubiak, Superlatt. Microstruct. 18, 275 (1995). 36. R. Osifichin, W. Mahoney, R. Andres, J. Henderson, C. Kubiak, and R. Dominey, Superlatt. Microstruct. 18, 283 (1995). 37. J. R. Blackborow and D. Young, “Metal Vapor Synthesis,” Springer-Verlag, New York, 1979. 38. S. Roginsky, A. Schalnikoff, Kolloid Z. 43, 67 (1927).

282 39. J. S. Bradley in “Clusters and Colloids: From Theory to Applications.” (G. Schmid , Ed.). VCH, Weinheim, 1994. 40. K. Kimoto, J. Phys. Soc. Jpn. 9, 762 (1953). 41. H. Morimoto, J. Phys. Soc. Jpn. 13, 1015 (1958). 42. K. Mihama and Y. Yasuda, J. Phys. Soc. Jpn. 21, 1166 (1966). 43. T. Hayashi, T. Ohno, S. Yatsuya, and R. Uyeda, Japanese J. App. Phys. 16, 705 (1977). 44. K. Kimoto and I. Nishida, Japanese J. App. Phys. 6, 1047 (1967). 45. K. Kimoto and I. Nishida, J. Phys. Soc. Jpn. 22, 940 (1967). 46. T. Okazaki, Japanese J. App. Phys. 27, 2037 (1988). 47. N. Wada, Japanese J. App. Phys. 8, 551 (1969). 48. R. Uyeda, in “Phys.-Math Society,” Japan, 1942. 49. A. Tasaki, S. Tomiyama, S. Iida, and R. Uyeda, Japanese J. App. Phys. 4, 707 (1965). 50. R. Uyeda, Progr. Mater. Sci. 35, 1 (1991). 51. J. Chang and J. Huang, Biotech. Prog. 14, 735 (1998). 52. D. W. Darnall, B. Greene, M. T. Henzl, M. J. Hosea, R. A. McPherson, and J. Sneddon, Environ. Sci. Technol. 20, 206 (1986). 53. D. W. Darnall, B. Greene, and J. L. Gardea-Torresdey, BioHydroMetall. 487 (1988). 54. A. V. Pethkar and K. M. Paknikar, J. Biotech. 63, 121 (1998). 55. N. Kuyucak and B. Volesky, Biorecovery 1, 189 (1989). 56. N. Kuyucak and B. Volesky, Precious Metals 211 (1986). 57. M. J. Hosea, B. Greene, R. A. McPherson, M. T. Henzl, M. D. Alexander, and D. W. Darnall, Inorgan. Chem. Acta 161 (1986). 58. J. A. Erdman, B. F. Leonard and D. M. Mckown, A case for Plants in Exploration: Gold in Douglas-fir at the Red Mountain Stockwell, U.S. Department of the Interior Geological Survey, Yellow Pine Distric, ID, 1985, p. 85. 59. K. C. Jones and P. J. Peterson, Biogeochem. 7, 3 (1989). 60. C. W. Anderson, R. R. Brooks, A. Ciarucci, C. J. LaCoste, M. Leblanc, B. H. Brobinson, R. Simcock, and R. B. Stewart, J. Geochem. Explor. 67, 407 (1999). 61. A. L. Kovalebskii and O. M. Kovalevskaya, Appl. Geochem. 4, 369 (1989). 62. J. R. Lujan, D. W. Darnall, P. C. Stark, G. D. Rayson, and J. L. Gardea-Torresdey, Solvent Extract. Ion Exchange 12, 803 (1994). 63. J. L. Gardea-Torresdey, K. J. Tiemann, J. G. Parsons, G. Gamez, and M. Yacamán, Adv. Environmental Res. 00, 1 (2001). 64. J. R. Peralta, J. L. Gardea-Torresdeay, K. J. Tiemann, E. Gómez, S. Arteaga, E. Rascón, and J. G. Parsons, Bull. Environ. Contam. Toxicol. 66, 727 (2001). 65. J. L. Gardea-Torresdey, J. G. Parsons, E. Gómez, J. Peralta-Videa, H. Troiani, P. Santiago, and M. J. Yacamán, Nano Lett. 2, 397 (2002). 66. M. J. Hostetler, J. E. Wingate, C. J. Zhong, J. E. Harris, R. W. Vachet, R. M. Clark, J. D. Londono, S. J. Green, J. J. Stokes, G. D. Wignall, G. L. Glish, M. D. Porter, N. D. Evans, and R. W. Murray, Langmuir 14, 17 (1998). 67. L. A. Porter, Jr. D. Ji, S. Westcott, M. Graupe, R. Czernuszewicz, N. Halas, and T. Lee, Langmuir 14, 7378 (1998). 68. D. V. Leff, P. C. Ohara, J. R. Heat, and W. M. Gelbart, J. Phys. Chem. 99, 7036 (1995). 69. M. Hostetler, S. Green, J. Stokes, and R. Murray, J. Am. Chem. Soc. 118, 4212 (1996). 70. M. Alvarez, J. Khoury, T. Schaaff, M. Shafigullin, I. Vezmar, R. L. Whetten, Chem. Phys. Lett. 266, 91 (1997). 71. R. S. Ingram, M. Hostetler, and R. Murray, J. Am. Chem. Soc. 119, 9175 (1997). 72. P. Buning, B. Humbel, A. Philipse, and A. Verklej, Langmuir 18, 3921 (1997).

Crystallography and Shape of Nanoparticles and Clusters 73. S. Chen and K. Kimura, Langmuir 15, 686 (1999). 74. S. R. Johnson, S. D. Evans, S. Mahon, and A. Ulman, Langmuir 13, 51 (1997). 75. M. Brust, D. Bethell, D. Schiffrin, C. Kiely, and L. Novel, Adv. Mater. 7, 795 (1995). 76. A. Templeton, M. Hostetler, C. Craft, and R. Murray, J. Am. Chem. Soc. 120, 1906 (1998). 77. I. Srnová-loufová, F. Lednick, A. Gemperle, and J. Gemperlová, Langmuir 16, 9928 (2000). 78. M. J. Yacamán, K, Heinemann, and H. Popa, Crit. Rev. Sol. Stat. Mater. 10, 243. (1983). 79. K. Yagi, “High resolution Electron Microscopy and Associated Techniques,” Oxford, 1988, p. 568. 80. K. Yagi, K. Takanayagi, K. Kobayashi, and G. Honjo, J. Cryst. Growth 28, 117 (1975). 81. M. José-Yacamán, J. A. Ascencio, H. B. Liu, and J. L. GardeaTorresdey, J. Vac. Sci. Technol. B 19, 1091 (2001). 82. Z. L. Wang, “Characterization of Nanophase Materials.” WileyVCH, Weinheim, 2000. 83. S. A. Harfenist, Z. L. Wang, R. L. Whetten, I. Vezmar, and M. Alvarez, Adv Mater. 9, 817 (1997). 84. J. Spiller Metois and J. A. Venables, Philos. Mag. 46, 1015 (1982). 85. M. Gillet, Surf. Sci. 67, 139 (1977). 86. L. D. Marks, “The structure of small silver particles.” Cambridge University Press, Oxford, 1980. 87. M. J. Yacamán and J. M. Domínguez, J. Catal. 213 (1980). 88. Z. L. Wang, T. S. Ahmad, and M. A. El-Sayed, Surf. Sci. 380, 302 (1996). 89. M. José-Yacamán, J. A. Ascencio, H. B. Liu and J. GardeaTorresday, J. Vac. Sci. Technol. B 19, 1091 (2001). 90. S. Giorgio and J. Urban, Appl. Phys. Lett. 52, 1467 (1988). 91. See “Nanoscale Characterization of Materials,” MRS Bulletin 22, 8 (1997). 92. “Handbook of Nanostructured Materials and Nanotechnology.” (H. S. Nalwa, Ed.). Vols. 1–5, Academic Press, London, 1999. 93. J. A. Ascencio, C. Gutierrez-Wing, M. E. Espinosa, M. Marín, S. Tehuacanero, C. Zorrilla, and M. José-Yacamán, Surf. Sci. 396, 349 (1998). 94. E. J. Kirkland, “Advanced Computing in Electron Microscopy.” Plenum, New York, 1998. 95. Z. L. Wang, “Elastic and Inelastic Scattering in Electron Diffraction and Imaging.” Plenum, New York, 1995. 96. L. D. Marks, Rep. Prog. Phys. 57, 603 (1994). 97. “Extended Icosahedral Structures.” (M. V. Jaric and D. Gratias, Eds.). Academic, 1987. 98. O. Krivanek, in “High Resolution Electron Microscopy.” (P. Busec, J. Cowley, and L. Eyring, Eds.). Oxford University Press, 1988. 99. J. M. Montejano-Carrizales, Institute of Physics University of San Luis Potosí, Internal report and personal communication, 2000. 100. L. D. Marks, J. Cryst. Growth 61, 556 (1984). 101. K. Koga and K. Sagawara, Surf. Sci. 529, 23 (2003). 102. J. Ascencio, M. Perez, and M. José-Yacamán, Surf. Sci. 447, 73 (2000). 103. Y. Chushak and L. S. Bartell, Eur. Phys. J. D 16, 43 (2001). 104. K. Manninen and M. Manninen, Eur. Phys. J. D 20, 243 (2002). 105. J. Uppenbrik and D. J. Wales, J. Chem. Phys. 96, 8520 (1992). 106. J. Urban, H. Sack-Kongehl, and K. Weiss, Z. Phys. D 28, 247 (1993). 107. D. Lu and K. Tanaka, Surf. Sci. 409, 283 (1998). 108. S. A. Nepijko, H. Hofmeister, H. Sack-Kongehl, and R. Schlogl, J. Cryst. Growth 312, 129 (2000).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Cyclodextrins and Molecular Encapsulation J. Szejtli Cyclolab Ltd., Budapest, Hungary

CONTENTS 1. 2. 3. 4. 5. 6.

Introduction Cyclodextrins Cyclodextrin Derivatives Inclusion Complexes Industrial Applications Cyclodextrin Based Nanostructures Glossary References

1. INTRODUCTION The interaction of two or more molecules which results in significant changes in their properties without establishing covalent chemical bond(s) between them is the subject of the supramolecular chemistry. Because one molecule is the possible smallest unit of any compound, the smallest nanostructure is a few nanometer sized molecular structure. A larger cylinder-shaped molecule, which is able to include another smaller molecule (of adequate size and polarity) into its axial cavity itself is a capsule of molecular size. The temporary or static inclusion of other molecules into the cavity of a so-called “host” molecule without establishing covalent bond results in broadly and practically very well utilizable effects, which belong to the ultimate limits of the nanotechnology. The cyclodextrins are toroidal (conical cylinder) shaped hollow molecular structures; their outer diameter is 13.7 to 16.9 nanometers, height uniformly 7.8 nanometers, and their axial cavity about 0.57 to 0.95 nanometers wide. They are produced by a relatively simple enzymic conversion technology from starch. A cyclodextrin (CD) molecule is the possible smallest (nano)capsule: in most cases another, smaller molecule can be inserted into its axial cavity. This “molecular encapsulation” results in smaller or larger changes in the physical and chemical properties of the encapsulated molecules. ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

While in the 1970s this possibility was not more than a scientific curiosity, by the beginning of the 21st century it became a widely used industrial technology; cyclodextrins are produced and utilized in thousand ton amounts. The number of cyclodextrin containing products and cyclodextrin utilizing technologies increases steadily—the cyclodextrin market increases yearly by about 15 to 20%. By end the of 2002, the number of CD-related publications is more than 26.000, representing more than 170,000 printed pages [1–26]. While 30 years ago about four to five CD papers were published monthly, in 2002 that many are published daily. About 16% of all CD-relevant publications are dedicated to the fundamentals of cyclodextrin chemistry and technology (i.e., the physical and chemical properties of cyclodextrins, their enzymology, toxicology, production, and derivatives). This section also includes the numerous review articles on CDs. Nearly 22% of the publications are dedicated to studies of the CD-inclusion phenomena. These works are generally not directly practice-oriented, dealing with energetics and kinetics of inclusion, X-ray, Fourier transform infrared, liquid and solid phase nuclear magnetic resonance (NMR), electron paramagnetic resonance, circular dichroism, Raman spectroscopy, enhancement of luminescence and phosphorescence, thermal analysis, interaction of CDS with specific guest types, enzyme modelling with CDs and CD derivatives, preparation, analysis of cyclodextrin complexes, etc. These methods, as well as the correlation between the complexation and various structural and external parameters, form the basis for all practical applications of CDs. The largest group of CD papers, nearly 25%, is dedicated to the pharmaceutical application of CDs. The large number (nearly 5000) of drug/CD related papers and patents is a little misleading, because many authors publish the same result in different journals under different titles, but with virtually identical content. Rediscoveries are published frequently, simply because the authors did not read the earlier literature; in essence they have discovered something that was published earlier. Because of the many repetitions, and the

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (283–304)

Cyclodextrins and Molecular Encapsulation

284

HO

OH 4

CH

HO

H O CH 2O

HO

CH4OH

OH

O

HO

H

4

O

OH O

OH

2

CH

4

OH

H

CH

OH

βCD O HO

H O

2

O

CH2OH

C H O H2 O

CH2OH

O

O HO O H O

HO O H O

CH

OH

αCD

O

H

O H4

O

O

C

O

O

H

O HO

O

OH

H

O

O

O

CH4OH

4

O

CH2OH

HO O

OH

H

H

O

OH O OH O

C

O

O

HO

HO

O H

O H O

H

O

O

O HO

O

O H2 C

O HO

HO

O

O

O

OH CH 4 O

CH

H O H2 C

OH

H

O

H H O

O O H

OH O

O

OH

O

O

CH

OH CH2OH

CH

2

OH

O O

H

OH

O

CH2OH

O HO

OH O

O

2

nonfeasible ideas, only about 30% of the published papers disclose really new and significant results. Actually only 7% of the CD-related papers are dedicated directly to the food, cosmetic, and toiletry applications of CDs, but at the same time about 70% of all cyclodextrins produced are used in this field. The approval process for CD-containing products in this field is much simpler and faster than in the case of the drug/CD formulations. The amount of CD used in a single cosmetic or toiletry product might be larger, by orders of magnitude, than the total amount used in drugs. Presently, about 11% of the CD literature is dedicated to the application of CDs in the chemical and biotechnological industries. About 25% of the CD literature involves the application of cyclodextrins in analytical chemistry and diagnostic preparations. The analytical applications of CDs refer mainly to the application of cyclodextrins in gas chromatography, in high performance liquid chromatography, and in capillary zone electrophoresis, but some dozen papers are dedicated to thin-layer chromatography, to enhancement of ultraviolet-visible (UV-VIS) absorption, luminescence/phosphorescence by CDs, and to increasing the sensitivity of the related analytical methods. Apparently, it is difficult to find a separation problem on an analytical scale which could not be solved by using the appropriate CD.

γCD

2. CYCLODEXTRINS 2.1. Chemistry Cyclodextrins comprise a family of three well known industrially produced major and several rare, minor cyclic oligosaccharides [1, 4, 9, 19]. The three major cyclodextrins are crystalline, homogeneous, nonhygroscopic substances, which are toruslike macrorings built up from glucopyranose units (Fig. 1). The -cyclodextrin (Schardinger’s -dextrin, cyclomaltohexaose, cyclohexaglucan, cyclohexaamylose, CD, ACD, C6A) comprises six glucopyranose units, CD (Schardinger’s -dextrin, cyclomaltoheptaose, cycloheptaglucan, cycloheptaamylose, CD, BCD, C7A) comprises seven such units, and CD (Schardinger’s dextrin, cyclomaltooctaose, cyclooctaglucan, cyclooctaamylose, CD, GCD, C8A) comprises eight such units (Fig. 1). The most important characteristics of the CDs are summarized in Table 1. As a consequence of the 4 C1 conformation of the glucopyranose units, all secondary hydroxyl groups are situated on one of the two edges of the ring, whereas all the primary ones are placed on the other edge. The ring, in reality, is a cylinder or, better said, a conical cylinder, which is frequently characterized as a doughnut or wreathshaped truncated cone. The cavity is lined by the hydrogen atoms and the glycosidic oxygen bridges, respectively. The nonbonding electron pairs of the glycosidic oxygen bridges are directed toward the inside of the cavity producing a high electron density there and lending to it some Lewis-base characteristics. The C-2-OH group of one glucopyranoside unit can form a hydrogen bond with the C-3-OH group of the adjacent

Figure 1. Chemical structure of -, -, and CDs.

glucopyranose unit. In the CD molecule, a complete secondary belt is formed by these H-bonds; therefore the CD is a rather rigid structure. This intramolecular hydrogen bond formation is probably the explanation for the observation that CD has the lowest water solubility of all CDs. The hydrogen-bond belt is incomplete in the CD molecule, because one glucopyranose unit is in a distorted position. Consequently, instead of the six possible H-bonds, only four can be established simultaneously. The CD is a noncoplanar, more flexible structure; therefore it is the more soluble of the three CDs. Figure 2 shows a sketch of the characteristic structural features of CDs. On the side where the secondary hydroxyl groups are situated, the cavity is wider than on the other side where free rotation of the primary hydroxyls reduces the effective diameter of the cavity. The cyclodextrins are produced from starch by enzymic conversion [9, 17]. The starch might be of any origin: maize, potato, tapioca, etc. All starch consists of two essential components: the amylose (2 to 70%), which is a linear glucose polymer (several hundreds or thousands of d-glucopyranose units are linked by -1,4 glucosidic linkages), and the amylopectin, which is a ramified structure. The main linkage type between the d-glucopyranoside units is also of -1-4 type, but every 20th to 30th glucose is connected to another polyglucan chain by an -1,6-glucosidic linkage. Splitting the glucosidic linkage (catalyzed by acid, or enzyme) the formed free carbenium cation will react immediately with another hydroxyl-containing molecular entity. If

Cyclodextrins and Molecular Encapsulation

285

Table 1. Characteristics of -, -, and CD. Parameter No. of glucose Mol. wt. Solubility in water (g/100 ml−1 ) at room temp. D 25 Cavity diameter (Å) Height of torus (Å) Diameter of outer periphery (Å) Approx. volume of cavity (Å)3 Approx. cavity volume in 1 mol CD (ml) In 1 g CD (ml) Crystal forms (from water) Crystal water wt% Diffusion constants at 40  C Hydrolysis by A. oryzae -amylase pK (by potentiometry) at 25  C Partial molar volumes in solution (ml mol−1 ) Adiabatic compressibility in aqueous solutions (ml mol−1 bar−1 104 







6 972 14.5 150 ± 0.5 4.7–5.3 7.9 ± 0.1 14.6 ± 0.4 174 104 0.10 hexagonal plates 10.2 3.443 negligible 12.332 611.4 7.2

7 1135 1.85 162.5 ± 0.5 6.0–6.5 7.9 ± 0.1 15.4 ± 0.4 262 157 0.14 monoclinic parallelograms 13.2–14.5 3.224 slow 12.202 703.8 0.4

8 1297 23.2 177.4±0.5 7.5–8.3 7.9 ± 0.1 17.5±0.4 427 256 0.20 quadratic prisms 8.13–17.7 3.000 rapid 12.081 801.2 −50

the reaction partner is water, the reaction is called hydrolysis; the final product is glucose. If the reaction partner is, for example, methanol, the reaction product is methylglucopyranoside. If the reaction partner is the sixth, seventh, or eighth glucopyranose unit of the same amylose CAVITY VOLUME 174 Å3

262 Å3

427 Å3

molecule or amylopectin molecule side chain, the product is a cyclodextrin. Because the minimum energy conformation for both the amylose as well as the segments of the amylopectin molecule is the helical (wormlike) structure, one turn is composed of six, seven, or eight glucose units. Consequently the sixth, seventh, or eighth glucopyranose unit will be in the immediate vicinity of the newly formed carbenium cation. This “splitting and the ring closing” process is catalyzed by the so called cyclodextrin-glucosyl-transferase enzyme.

2.2. Production αCD

βCD

γCD

in one mol : 104 ml

157 ml

256 ml

0,14 ml

0,20 ml

in one g : 0,10 ml

Secondary OH-side Hydrophylic region

cross section

Hydrophobic region Primary OH-side

Figure 2. Dimensions and hydrophilic/hydrophobic regions of the CD molecules.

The cyclodextrin glucosyl transferase enzyme (CGT-ase) is produced by a large number of microorganisms, like Bacillus macerans, Klebsiella oxytoca, Bacillus circulans, Alkalophylic bacillus No. 38-2, etc. Genetic engineering has provided more active enzymes, and probably, in the future, mostly these enzymes will be used for industrial CD production [17]. The first step in CD production is the liquefaction of the starch at elevated temperature. To reduce the viscosity of the rather concentrated (around 30% dry weight) starch solution, it has to be hydrolyzed to an optimum degree. The prehydrolyzed starch must not contain glucose or low molecular oligosaccharides, because they strongly reduce the yield of the formed CDs. After cooling to the optimum temperature, CGT-ase enzyme is added to the starch solution. In the so-called nonsolvent technology, the formed -, -, and CDs have to be separated from the complicated partially hydrolyzed mixture. In the case of the solvent technology, an appropriate complex-forming agent is added to the conversion mixture. If toluene is added to this system, the formed toluene/CD complex is separated immediately, and the conversion is shifted toward the formation of CD (Fig. 3). If n-decanol is added to the conversion mixture, mainly CD will be produced, while in case of cyclohexadecenol the main product is CD. Various other complexforming agents can be used. The selection depends on price,

Cyclodextrins and Molecular Encapsulation

286

CH3 + maltodextrins CGT-ase

CH3

CH3

amylose

CH3

CH3

CH3

toluene/beta-CD-complex

Figure 3. Production of CD: in the presence of toluene the formed CD is converted immediately to a water-insoluble toluene/CD complex. Therefore the main product will be CD. In the presence of decanol CD, hexadecenol CD is the isolated final product.

toxicity, and explosivity, but mainly on the efficiency of the removal of the solvents from the crystalline end-product. The insoluble complexes are separated from the conversion mixture by filtration. The removal of the solvents from the filtered and washed complex is generally made after suspending it in water by distillation or extraction. The aqueous solution obtained after removal of the complexing solvent is treated with activated carbon and filtered. The cyclodextrins are then separated from this solution by crystallization and filtration. The homegeneity of the industrially produced cyclodextrins is generally better than 99%.

2.3. Biological Effects 2.3.1. Absorption and Metabolism For application of cyclodextrins in the pharmaceutical and food industry, the fate of cyclodextrins in the mammal organisms is a crucial question. Cyclodextrins are consumed by humans or animals (as orally administered pharmaceuticals, or as food additives) either as free cyclodextrins or their inclusion complexes, containing some drug, flavor, or other guest substance. The inclusion complexes undergo very rapid dissociation under physiological conditions in the gastrointestinal tract; therefore only the absorption of the free cyclodextrin deserves attention [8, 16, 17]. Summarizing the available data and experimental observations the following conclusions can be drawn. The CD molecule is a relatively large molecule. Its outer surface is strongly hydrophilic. It is a true carrier. It brings the hydrophobic guests into solution, keeps them in dissolved state, and transports them to the lipophilic cell membrane, but after delivering the guest to the cell (because it has higher affinity to the guest than the cyclodextrin), the cyclodextrin remains in the aqueous phase. Only an insignificant amount of orally administered CD is absorbed from the intestinal tract in intact form. The source of eventual CD-elicited toxic phenomena might be that CDs can solubilize and facilitate the otherwise insoluble, nonabsorbable toxic compounds, like polyaromatic hydrocarbons, pesticides, etc. The preponderant part of orally administered cyclodextrin is metabolized in the colon, by

the colon microflora. The primary metabolites (acyclic maltodextrins, maltose, and glucose) are then further metabolized, absorbed like the starch, and finally excreted as CO2 and H2 O. While the metabolism of starch takes place in the small intestines, the cyclodextrin is metabolized in the colon. Correspondingly the maximum intensities of metabolism are observed for starch around 1 to 2 hours and for CDs 6 to 8 hours after consumption, respectively. The metabolism of CD is slower and that of CD is much faster than that of CD. The chemically modified CDs are, however, very resistant to enzymic degradation. Methylated and hydroxypropylated CDs (and probably all other more or less substituted CD derivatives) can be absorbed from the intestinal tract to a quite considerable extent and can get into the circulation.

2.3.2. Toxicology Upon parenteral administration all CDs can interact with the components of cell membranes, but the extent of this interaction is very different. This cell-damaging effect can be characterized by CD concentration, which results in an 50% haemolysis of the erythrocytes or which reduces to 50% the luminescence of an Escherichia coli suspension. In these experiments, the more hydrophobic (but very well soluble) derivatives like dimethyl CD showed the largest effects. Because the cell-damaging effect is a consequence of CD complexation of cell-wall components like cholesterol and phospholipids, CD and particularly its more hydrophobic derivatives, like the dimethyl CD, show the strongest cellmembrane damaging effect among all CDs. The CD and its derivatives are less toxic, because their affinity (through the wider CD cavity) to cholesterol and phospholipids is lower. Administering the CDs at a rate that their actual concentration does not reach the haemolytic threshold concentration (slow infusion, or absorption from the intestinal tract, or through the skin), this immediate toxic consequence can be avoided, but not the kidney-toxic effect. Parenteral administration of CD is not possible, because it forms the most stable cholesterol/CD complex, which will be accumulated in the kidneys, destroying cells. Apparently CD is free from this noxious property: it is well soluble and does not form stable, easily crystallizable complexes with any vital component of the circulation. The cavity of CD is too large and that of the CD is too small for cholesterol. HPCD (hydroxypropyl CD) does not form insoluble kidney-damaging cholesterol complex crystals; therefore it can be injected even in rather large doses without any irreversible consequences, but chronic treatment again showed unwanted side-effects: it forms soluble cholesterol/HPCD complex in the circulation. In the kidneys this less stable complex dissociates. The HPCD will be excreted, but the cholesterol remains in the kidneys. Long-term parenteral administration of HPCD will heavily load the kidneys with cholesterol—not considering other toxic effects of HPCD. Parenterally only such CDs can be administered that have a low affinity to cholesterol. The best CD would be the one that is stable in aqueous solution, has a high drug solubilizing and stabilizing capacity, but after injecting into the circulation rapidly decomposes. Regrettably we have not yet such a derivative.

Cyclodextrins and Molecular Encapsulation

287

3. CYCLODEXTRIN DERIVATIVES

O

OH

HO

2

HO

CH HO CH 2 O O

OH O OH

O

OH CH

OH CH CH 3

CH 2 OH

e

H 3 O CH CH H2 OC O

2

O CH2OMe O OH O Me O M eO

O HO

M

CH

O

2

CH2OH

OH

Me O

CH

2

OH HO

O O Me HO

O HO HO

O HO

HO O HO

O

e M O H2 O

O

O

O

CH2OMe O

C

O

OH OH

MeO

eO

OCH 2CHCH 3 CH CH 2OCH 2CH 3 O

O

OM O e

2

O HO

O

From the thousands of CD derivatives described in hundreds of scientific papers and patents only a few can be taken into consideration for industrial scale synthesis and utilization. Complicated multistep reactions, using expensive, toxic, environment-polluting reagents and requiring purification of the products by chromatography, are feasible to prepare derivatives only on laboratory scale. To produce tons at an acceptable price only about a dozen of the known CD derivatives can be taken into consideration. Among industrially produced, standardized, and available (even in ton amounts) CD derivatives the most important ones are the heterogeneous, amorphous, highly water soluble methylated CD and 2-hydroxypropylated CDs (Fig. 4). Due to its heterogeneity the HPCD cannot be crystallized, which is an important advantage (e.g., at producing liquid drug formulations). The first hydroxypropyl-CD and hydroxypropyl-CD containing drug formulations are already approved and marketed in several countries. A methylated CD is more hydrophobic than the CD itself; therefore it forms a more stable (but soluble) complex with cholesterol. Its affinity to cholesterol is so strong that it extracts cholesterol from the blood cell membranes, resulting in hemolysis already at around 1 mg/mL concentration.

CH M

2

3.2. Derivate Types

O O Me

H

• to improve the solubility of the CD derivative (and its complexes) • to improve the fitting and/or the association between the CD and its guest, with concomitant stabilization of the guest, reducing its reactivity and mobility • to attach specific (catalytic) groups to the binding site (e.g., in enzyme modeling), • to form insoluble, immobilized CD-containing structures, polymers, (e.g. for chromatographic purposes).

O HO

C

For a series of reasons (price, availability, approval status, cavity dimensions, etc.) the CD is the most widely used and represents at least 95% of all presently produced and consumed CDs. It is used for many purposes; however, its anomalous low aqueous solubility (and the low solubility of most of its complexes) is a serious barrier in its wider utilization. Fortunately by chemical or enzymatic modifications the solubility of all CDs can be strongly improved, and instead of the 18 g/L aqueous CD solutions 500 g (or more)/L aqueous CD derivative solutions can easily be prepared. In the cyclodextrins every glucopyranose unit has 3 free hydroxyl groups which differ in both their functions and reactivity. The relative reactivities of C(2) and C(3) secondary and C(6) primary hydroxyls depend on the reaction conditions (pH, temperature, reagents). In CD 21 hydroxyl groups can be modified substituting the hydrogen atom or the hydroxyl group with a large variety of substituting groups like alkyl-, hydroxyalkyl-, carboxyalkyl-, amino-, thio-, tosyl-, glucosyl-, maltosyl-, etc. groups, thousands of ethers, esters, anhydro-deoxy-, acidic, basic, etc. derivatives can be prepared by chemical or enzymatic reactions. The aim of such derivatizations may be:

O

CH2CHCH3

3.1. Aim of Derivatization

O H

CH 2 OMe

OH e OM CH 2 O

Figure 4. Structure of the heptakis (2,6-O-dimethyl)-CD and of a hydroxypropyl CD of DS∼4.

A particular methylated CD, the heptakis (2,6-di-Omethyl-CD) is a crystalline product (Fig. 4). It is extremely soluble in cold water but insoluble in hot water. Therefore, its purification, and also the isolation, of its complexes is technically very simple. Up to now no better solubilizer has been found among the CDs. It is available in better than 95% isomeric purity for injectable drug formulation (for example radio-pharmaceuticals) but for widespread industrial application the cheaper noncrystalizable randomly methylated CD (called RAMEB) is produced and marketed. Enzymatically modified CDs are produced, for example, by reacting CD with starch in the presence of pullulanase enzyme. One or two maltosyl or glucosyl groups will be attached to the primary side of the CD ring with -1,6 glycosidic linkage. The product is the so-called “branched” CD (mono- or dimaltosyl or mono- or diglucosyl CD) which is very well soluble in water, being a heterogeneous, noncrystallizable substance. It is produced and used in the food industry, mainly in Japan, for example, for production of stable flavor powders. There has been increasing interest toward the per-acylCDs. All acetylated CDs from per-acetyl to per-octanoyl esters have been studied partly as retarding drug carriers and partly as bioadhesive, film-forming substances to be used in transdermal drug delivery systems. The heptakis(sulfobutyl)-CD is very well soluble in water, is noncrystallizable, and even at extreme high doses seems to be free from any toxic side-effects. It can be used as a chiral separating agent in capillary zone electrophoresis, but the aim of the intensive research is to use it as a parenteral drug carrier, for preparation of aqueous injectable solutions of poorly soluble drugs. A few such drug complexes are already approved and marketed. The CD sulfates possess many similar properties as the heparin without its anticoagulating effect. Apparently they can reduce the blood supply of tumor tissues through inhibiting the formation of new arteries (=antiangiogenesis). The monochloro-triazinyl CD is produced on industrial scale from CDs and cyanuric chloride. It is reactive with cellulose fibers in alkaline medium (see at the application of CDs in textiles). To elongate the actual CD cavity substituents are attached to the primary or secondary side. This elongation may be hydrophylic in which case hydroxyalkyl groups are attached

Cyclodextrins and Molecular Encapsulation

288

CH3

CH3

CH3

CH3

Figure 5. Schematic representation of the association–dissociation of the host (CD) and guest (p-xylene). The formed guest/host inclusion complex can be isolated as a microcrystalline powder.

The formed inclusion complexes can be isolated as stable crystalline substances. Upon dissolving these complexes, an equilibrium is established between dissociated and associated species, and this is expressed by the complex stability constant Ka . The association of the CD and guest (D) molecules, and the dissociation of the formed CD/guest complex is governed by a thermodynamic equilibrium: CD + D ⇋ CD · D K11 =

CD · D CDD

Generally, the study of the interaction between a CD and a potential guest begins with registration of the solubility isotherm (phase solubility diagram, Fig. 6). Adding the potential “guest” to an aquous solution and agitating (stirring, shaking) until the equilibrium is attained AP AL

AN

[Dissolved guest]St

to the ring, or it might be hydrophobic. For example substituting the primary hydroxyl groups with long fatty-acid chains, “medusa”-like molecules can be prepared. These molecules behave as detergents while retaining their complex forming ability. The coming years will decide how utilizable these derivatives are. The chair conformation of the CD ring can be modified by inverting the position of some hydroxyl groups. For example eliminating the tosyl group in alkaline medium from a CD-tosyl derivative, 2,3-anhydro derivatives can be prepared. Opening the anhydro ring one hydroxyl group will take up an inverted position, and this way cycloaltrins are formed. Eliminating an appropriate leaving group from the primary side, 3,6-anhydro-CDs are formed. Because of the twisted conformation of the anhydro glucopyranose unit the properties of CDs—for example, their solubility—are strongly increased. It is possible to close one side of the CD cavity, for example, by overbridging the primary or secondary side with an appropriate bifunctional substituent. It is expected that these overbridged CDs will form more stable complexes with certain guests. A dozen various CD derivatives are used in gas chromatography, liquid chromatography, and capillary zone electrophoresis. For other industrial purposes where toxicological demands do not play a decisive role, epichlorohydrin crosslinked, hydroxyethylated, or sometimes (apparently rather fancy, but by its use justified) mixed ethers–esters, like heptakis (2,6-di-O-methyl)-3-O-trifluoracetyl-CD and similar derivatives, are produced and utilized. It seems to be very probable that for drug carrier purposes four or five different CD derivatives will be developed and produced in the future, because no one of them is able to fulfill alone all the very strict requirements, which are usual in the case of a parenteral drug carrier. For other industrial purposes the production of hundreds of tons of alkylated, hydroxyalkylated, and acylated CDs is prognosticated by end of first decade of the 21st century.

A

B

4. INCLUSION COMPLEXES 4.1. Mechanism and Equilibria In an aqueous solution, the slightly apolar cyclodextrin cavity is occupied by water molecules which are energetically unfavored (polar–apolar interaction) and therefore can be readily substituted by appropriate “guest molecules” which are less polar than water (Fig. 5). The dissolved cyclodextrin is the “host” molecule, and the “driving force” of the complex formation is the substitution of the high-enthalpy water molecules by an appropriate “guest” molecule. One, two, or three cyclodextrin molecules contain one or more entrapped “guest” molecules. Most frequently the host:guest ratio is 1:1. This is the essence of “molecular encapsulation” [17]. This is the simplest and most frequent case. However, 2:1, 1:2, 2:2, or even more complicated associations and higher order equilibria exist, almost always simultaneously.

S0 Bs

Sc BI

[CD]

Figure 6. Registering the solubility of a poorly water soluble guest as a function of the CD concentration in aqueous solution, different solubility isotherms can be obtained, depending on the type of CD and properties of the guest. Solubility isotherm types: So = solubility of guest in the absence of CD; St = concentration of dissolved guest (free + complexed); and Sc = solubility limit of the poorly soluble complex. Isotherm types Ap , AL , and AN = very soluble complexes formed (solubility limit determined by solubility of the CD). Isotherm BS = complex of limited solubility is formed. Isotherm BI = insoluble complex is formed.

Cyclodextrins and Molecular Encapsulation

(from 3 h to 1), the following visual observations may be expected: If the guest is well soluble in water then • No visual change, but a shift in the UV spectrum, a change in the NMR chemical shift values, enhanced fluorescence, or induced circular dichroism, etc. may recorded. • Precipitate is formed already at low concentrations of the guest which allows one to conclude to the complex formation. (The majority of water miscible solvents at high concentration precipitate the CDs, but it does not necessarily mean complex formation.) • Solubility of the CD increases significantly (e.g., solubility of CD increases from 18 to over 100 mg/ml) which points to a strong interaction with the CD, but not necessarily inclusion. If the guest is poorly soluble in water, and it is able to form complex with the CD, then its solubility as a function of the CD concentration in most cases will change. This change may be an increase, either monotonously or only to a certain limit, or it may even decrease. The correlation between guest solubility and CD concentration is illustrated by Figure 6. If only dissolved complex is formed the phase-solubility isotherm is of “A” type, whereas if the solubility of the formed complex is limited, then the isotherm is of the “B” type. In cases where the formed complex is insoluble (type BI , Fig. 6) the solubility of the guest remains unaltered until all guest molecules are converted into insoluble complexes. Thereafter the concentration of dissolved guest begins to decrease. When the complex is more soluble than the free guest, but its solubility limit can be reached within the studied CD concentration range, the (St ) first increases from the aqueous solubility of the guest (SO ) until point “A,” where the solubility limit of the complex is reached. Any further increase in the CD concentration results in no further increase in solubility, but in precipitation of the microcrystalline complex (BS -type isotherm). Reaching point “B” means that all solid guest has been converted to a less soluble inclusion complex, therefore adding even more CD to the system. The association–dissociation equilibrium will be shifted to association, and the solubility asymptotically approximates to the inherent solubility limit of the complex (Sc ). Theoretically the concentration increase from So to A should be identical with the one from zero to Sc (i.e., at point “A” the solution is just saturated for both the guest and its complex). This is, however, only a theoretical case, because in many instances the stoichiometry of a formed complex depends on the concentration ratios. While at the beginning only a 1:1 complex is formed, at higher CD concentration the stoichiometry is more complicated (1:2, 2:3, etc.). The prevailing complex structure and stoichiometry are not necessarily identical in solution and in the solid state. If within the studied CD-concentration range the solubility limit of the complex is not reached, the isotherm is of type “A,” and “AL ” means a linear increase with unchanged stoichiometry. The “AP ”-type isotherm shows a

289 positive deviation from linearity and this indicates a continuous increase in the stoichiometry of the complex, (i.e., the original 1:1 complex tends to associate with further guests, forming 2:3, etc. compositions). If the isotherm is of “AN ” type, the system is even more complicated because it can point either to an increase of the host ratio within the complex (1:1 to 2:1) or a change in the solute-solvent interaction (hydration, ionization of the guest) or both. The determination of the complex stability constants and the problems associated with their interpretation are treated in Section 6. The theoretically incorrect, but for practical purposes utilizable, apparent complex stability constant KC (association constant) for a 1:1 complex can be calculated from the slope and intercept of the initial straight line portion of the diagram as follows: Kc

tan  St − So  = So CDt − St − So  So 1 − tan 

The stoichiometry of the inclusion compound can be deduced by analyzing the length of the plateau region, AB, according to the following equation: guest CD =

(total guest added to system) − guest in solution at point A CDt corresponding to plateau region

This equation means that the concentration of CD, corresponding to the plateau, is equal to that of the host which is consumed for the conversion to the solid inclusion compound, with no free solid guest remaining at point A. The stoichiometry thus estimated by this method should be confirmed by isolation and chemical analysis of the solid inclusion compound. In the case of the B1 -type diagram, the initial rise is not detectable because the inclusion compound is practically insoluble. In the same way, the Kc value is calculated from the AL -type diagram according to the first equation. From the Ap -type diagram Kc is calculated by iteration. The value of Kc cannot be calculated from an AN -type diagram.

4.2. Stoichiometry and Crystal Structure Geometrical rather than chemical factors are decisive in determining the kind of guest molecules which can penetrate into the CD cavity [9, 17]. The -, -, and CDs, with different internal diameters, are able to accommodate molecules of different size. Naphthalene is too bulky for CD, and anthracene fits only into CD. Propionic acid is compatible with CD, but it has no satisfactory fitting in the large cavities. (Fig. 7). The included molecules are normally oriented in the host in such a position as to achieve the maximum contact between the hydrophobic part of the guest and the CD cavity. The hydrophilic part of the guest molecule remains, as far as possible, at the outer face of the complex. This ensures maximum contact with both the solvent and the hydroxyl groups of the host. Complex formation with molecules significantly larger than the cavity may also be possible. This is done in such a way that only certain groups or side chains penetrate into

Cyclodextrins and Molecular Encapsulation

290 COOH CH3

A

B

NH

C

COOH

D O

COOH

OH

OH

OH

COOH

COOH OH

OH

OH

OH

F

E αCD K = 250 M-1

OH

OH

G βCD K = 1240 M-1

γCD

Figure 7. Examples of CD complexes: the same CD with different guests, and the same guest with different CDs. Toluene/CD (A), diphenylamine/CD (B), long-chain fatty acid/CD (C), short chain fatty acid + diethyl ether ternary CD complex (D), prostaglandin E2/CD (E), prostaglandin E2/CD complex (F), prostaglandin E2/CD (G) [9].

the carbohydrate cavity (e.g., aromatic amino acid moieties of a polypeptide or protein). The fitting between host and guest can be visualized by computer graphics. Computer programs are available which, using crystallographically determined bond length and bond angle values and appropriate interaction energy functions, can rapidly ascertain the most probable conformations of a complex. However, if an inclusion can be established on the screen of a computer monitor, it does not necessarily mean that the complex can really be prepared, because the crystallographic data cannot take into account those factors that exist only in aqueous solutions: hydration, ionization, and concomitant torsion of the conformation. A 2:1 complex of CD and a guest molecule may be formed when the guest is too large (long) to find complete accommodation in one cavity, and its other end is also amenable to complex formation. Such complexes are formed in the case of the vitamin D3/ CD complex and the much “slimmer” aliphatic chains of n-alkanes and CD. In solution both the - and CD are highly associated with benzoic acid, but only the CD complex can be readily crystallized. A crystalline 2:1 complex with CD and benzoic acid will give a low yield. CD will perfectly include a benzoic acid molecule; the carboxyl group forms a hydrogen bond with one secondary CD hydroxyl. The role of substituents in complex formation is demonstrated by the methylnitrophenols. Relative to 4-nitrophenol, methyl groups in positions 2 and 6 have no significant influence on the stability of the CD complex, but one methyl group in the 3-position will lower the stability of the complex by about two orders of magnitude. 3,5-Dimethyl-4nitrophenol fails to give a complex at all. Strongly hydrophilic molecules (very soluble in water) and strongly hydrated and ionized groups are not or are only very weakly complexable. Only molecules which are less polar than water can be complexed by CDs.

Excessively strong cohesive forces between the molecules of the guest impede their separation, which is a precondition for the inclusion. A measure of the cohesion between the molecules of a crystalline substance is the melting point. When the melting point is higher than 250  C a stable CD complex cannot generally be prepared. The structure of crystalline CD complexes is not necessarily identical to that of the complexes in solution. In the dissolved state the guest molecule or its corresponding group is located within the cavity of the CD and the whole complex is surrounded by a multilayer hydrate shell. In the crystalline state, however, guest molecules and also some CD molecules include only water molecules; consequently they are incorporated into the crystal lattice as water complexes. Therefore, the isolated amorphous or crystalline complexes are practically never of strictly stoichiometric composition; however, they are stable even if the ring cavities are only partially saturated by apolar guest molecules. In solution the association–dissociation equilibrium determines the uncomplexed/complexed ratio for the guest and the host. However, depending on properties such as size and shape of the guest, not only 1:1, but also 1:2, 2:1, or 2:2 host:guest complexes may coexist in solution. This will also depend on the CD and/or guest concentration. Even two identical or different guest molecules may be included in one host (ternary complex). For example sodium 1-pyrenesulphonate with CD forms a 1:1 low stability (k = 1 M−1  complex, but with CD a very stable (k = 106 M−1  2:1 complex is formed. At higher CD concentrations, however, one further CD will react, resulting in a 2:2 complex. Similarly, CD forms a 1:1 complex with methyl orange in dilute solution; depending on the concentration 1:2 and 2:2 complexes can also occur. Generally, in aqueous solution with no extreme concentrations, the 1:1 complex is predominant. CD inclusion complexes always contain water as crystal hydrate. For example, the p-iodoaniline/CD complex contains four and the krypton and the methanol/CD complexes contain five molcules of water of crystallization. In crystalline CD inclusion complexes, the packing depends largely on the type of guest molecule. The cage-type structure (see Fig. 8) is characteristic for small guests, which fit completely into the cavity. The channel type is observed when the guest is so large that it protrudes on both sides of the cavity. The structure of the complex changes from the cage to channel type in the homologous series of carboxylic acids. Acetic, propionic, and butyric acids form cage-type structures, but valeric acid is too long to be accommodated in the cage, and from valeric acid onward, only channel-type structures are formed. The ionization of a small guest also results in the conversions of the cage to channel structure. Thus, acetic acid forms a cage-type structure, whereas the ionic potassium acetate prefers a channel-type arrangement. The brickwork-type cage structure is characteristic of small, neutral, but not fully incorporated guests, which only weakly protrude on one side of the cavity (e.g., monocyclic aromatic compounds). This kind of brick-type cage arrangement has so far only been observed with - and not with - and CD.

Cyclodextrins and Molecular Encapsulation

291

A

B

C

CHANNEL

HERRINGBONE

BRICK

E

D

Figure 8. Channel (A), herringbone (B), brick-wall (C), head-to-tail (D), and head-to-head (E) alignment of CD rings in their complex crystal structures.

The CD complexes crystallize with a fishbone, dimeric cage or channel structure. The typical fishbone structure has been observed only with a few small guests: water, methanol, and ethanol. Larger guests, such as n-propanol, are accommodated in cavities which are formed by two CD molecules arranged in a head-to-head fashion, although n-propanol is small enough to be enclosed in a monomeric cavity. It appears that the head-to-head dimer formation is favored by CD. These “double volume cages” (i.e. head-to-head arranged dimers or even tetramers) are arranged side-by-side in layers. Such adjacent layers are displaced so that the cavities of the dimers or tetramers are closed at both ends, or they are stacked on top of each other. This leads to cage-type structure formation, although the CD molecules are arranged in a channel-type form. In fact, all the channels formed by CD are more or less irregular because this molecule has seven glucose units in one ring. If CD is crystallized as an “empty” molecule from water, it arranges in a fishbone-type cage structure similar to that observed for - and CDs. If, however, a small guest, such as n-propanol, is added, a channel-type structure is formed in which the CD molecules, with an eightfold symmetry, are stacked along a fourfold symmetry axis, and exactly linear channels are produced.

4.3. Effects of Inclusion The most important primary consequences of the interaction between a poorly soluble guest and a CD in aqueous solution are as follows: (a) The concentration of the guest in the dissolved phase increases significantly, while the concentration of the dissolved CD decreases. This latter point is not always true, however, because ionized guests, or hydrogenbond establishing (e.g., phenolic) compounds may enhance the solubility of the CD.

(b) The spectral properties of the guest are modified. For example, the chemical shifts of the anisotropically shielded atoms are modified in the NMR spectra. Also, when achiral guests are inserted into the chiral CD cavity, they become optically active and show strong induced Cotton effects on the circular dichroism spectra. Sometimes the maximum of the UV spectra are shifted by several nm and fluorescence is very strongly improved, because the fluorescing molecule is transferred from the aqueous milieu into an apolar surrounding. (c) The reactivity of the included molecule is modified. In most cases the reactivity decreases (i.e., the guest is stabilized), but in many cases the CD behaves as an artificial enzyme, accelerating various reactions and modifying the reaction pathway. (d) The diffusion and volatility (in case of volatile substances) of the included guest decreases strongly. (e) The formerly hydrophobic guest, upon complexation, becomes hydrophilic; therefore its chromatographic mobility is also modified. And in the solid state: (f) The complexed substance is molecularly dispersed in a carbohydrate matrix forming a microcrystalline or amorphous powder, even with gaseous guest molecules. (g) The complexed substance is effectively protected against any type of reaction, except that with the CD hydroxyls, or reactions catalyzed by them. (h) Sublimation and volatility are reduced to a very low level. (i) The complex is hydrophilic, easily wettable, and rapidly soluble. When, in an aqueous system, the formation of the CDinclusion complex can be detected (e.g., by NMR or circular dichroism, or through a catalytic effect), this does not mean necessarily that a well-defined crystalline inclusion complex can be isolated. The two main components of the driving force of the inclusion process are the repulsive forces between the included water molecules and the apolar CD cavity, on the one hand, and between the bulk water and the apolar guest, on the other hand. This second factor does not exist in the crystalline (dry) state. Therefore it is not uncommon that the complex formation is convincingly proven in solution, but the isolated product is nothing other than a very fine dispersion of the CD and the guest. The absolute majority of all practical applications of CDs are based on their inclusion complex forming capacity.

4.4. Preparation of CD Complexes The preparation of cyclodextrin inclusion complexes is simple; however, the conditions have to be “tailor made” for any guest substances [2, 4, 17]. The complexation may be performed in homogeneous solution, or in a suspension, under pressure, or by simple mixing of the components, by melting together the potential guest with the CD. The principle is illustrated in Figure 9. When preparing a CD complex in solution the presence of water is absolutely necessary. Either pure water or some

Cyclodextrins and Molecular Encapsulation

292 in water hydrophobic poorly soluble drug

dissolved CD

inclusion complex

be substituted by the guests. Besides this, the crystal lattice of the complex is also different from that of the water– CD complex. Therefore the reaction runs toward inclusion. Owing to their different crystal structure, the molecular layers of the inclusion complex formed on the surface of the parent cyclodextrin will dissociate from the crystal. Thus the entire cyclodextrin crystal quickly becomes disorganized and transformed into the guest–CD complex. Industrially the solution or slurry is freeze-dried in the case of expensive and very sensitive guests, like prostaglandins, vaccines, and taxanes, or spray-died like omeprazole, or simply kneaded and dried in air or in vacuo like nitroglycerin, essential oils, etc.

isolation

5. INDUSTRIAL APPLICATIONS crystalline complex

Figure 9. Schematic representation of the production of a CD complex.

aqueous system containing an organic solvent can be used. The use of organic solvent is necessary when the guest molecule is hydrophobic, or its melting point is over 100  C. Therefore it cannot be dispersed finely in an aqueous CD solution. In such a case, organic solvents have to be used for solubilization of the guest. Only a very limited number of organic solvents can be used, because the majority of them are excellent complexforming partners. Using ethanol, at least a small amount of it will be retained very firmly in the formed complex (seldom more than 1–2% by weight of the product). Other solvents, like diethyl ether, do not form a stable complex with CD; however, together with, for example, prostaglandin-F2 , a small amount of a ternary complex is formed, and then the diethyl ether cannot be removed, even at 105  C in vacuo. Complex for gases (xenon, chlorine, ethylene, krypton, carbon, dioxide, etc.) can be prepared by exposing a saturated CD solution to the gas at 7–120 atm for 5–8 days at 20  C. In the so-called “slurry” method the CD and the guest are not dissolved but only finely suspended in water at ambient temperature with vigorous stirring. (The use of ultrasonification may even improve the procedure by accelerating the dispersion of solid phase.) The reaction mixture is stirred intensively for 4–8 h (generally in the case of light oils, e.g., essential oils, terpenoids), or even for several days (24–72 h) in the case of heavy oils (e.g., natural waxes, higher terpenoids, balms) or when the guest substance is solid and no solvent is used. This method is the most feasible for industrial purposes. A variation of this method uses even less water. In the so-called “kneading” method the cyclodextrin is intensively kneaded with a small amount of water to which the calculated amount of the guest component is added directly without using any solvent. Because the initial cyclodextrin–water complex is energetically less favored than the CD–guest complex, the cavity water molecules will

5.1. Pharmaceutical Industry The complexation of a drug molecule with a CD should be taken into consideration when the bioavailability and/or the chemical stability of the drug is not satisfactory, or the usual formulation methods do not result in an acceptable product. Indications for CD complexation include: • The bioavailability of the drug (upon oral, dermal, pulmonar, mucosal, etc. application) is incomplete or irregular, because the drug is poorly soluble. • The rate of dissolution is low; even in case of a complete absorption, the time to reach the effective blood level of the orally administered drug is too long. • The drug is chemically unstable (on account of its autodecomposition, polymerization, or degradation by atmospheric oxygen, absorbed humidity, light, etc.). No marketable formulation with satisfactory shelf-life can be produced. • The drug is physically unstable. Volatilization or sublimation result in losses. By migration the originally homogeneous product becomes heterogeneous. On account of its hygroscopicity it liquifies by atmospheric humidity. • Because of the low solubility no aqueous eye-drop or injectable solution (or other liquid formulation) can be prepared. • The drug is a liquid, but its preferred pharmaceutical form would be a stable tablet, powder, aqueous spray, etc. • The drug is incompatible with other components of the formulation. • The dose of the lipid(like) hardly homogenizable drug is extremely low. Therefore content uniformity of the product is problematic. • The acceptability of the drug is bad, because of bad smell, bitter, astringent, or irritating taste. • Relief of serious side-effects (throat, eye, skin, or stomach irritation) is required. • Because of the extreme high biological activity (in case of drugs of extremely low doses) working with such powder is rather dangerous.

Cyclodextrins and Molecular Encapsulation

293

The advantageous results of CD complexation of (CDcomplexable) drugs are as follows: • improved bioavailability from solid or semisolid formulations (Fig. 10), • enhanced stability, elongated shelf-life, • reduced side-effects, • uniform, easy to handle powders, even from liquids, and • aqueous, injectable solution from poorly soluble drugs. Speaking only of the numerous advantages of drug/CD complexation can be bitterly misleading, because there are just as many limiting factors which restrict the applicability of CDs to certain types of drugs, because not all drugs are suitable for CD complexation. Many compounds cannot be complexed, or complexation results in no essential advantages. Inorganic compounds generally are not suitable for CD complexation. Exceptions are nondissociated acids (HCl, HI, H3 PO4 , etc.) halogens, and gases (CO2 , C2 H4 , Kr, Xe, etc.). Inorganic salts as KCl, Fe salts, etc. can not be complexed. General preconditions (not without exceptions!) to form a medicinally useful CD complex of a drug molecule are the following: • More than five atoms (C, P, S, N) form the skeleton of the drug molecule. • Its solubility in water is less than 10 mg/ml. In vitro dissolution

[dissolved drug]

drug/CD

drug

time In vivo absorption

[drug in blood]

drug/CD

drug

time

Figure 10. The bioavailability of a poorly soluble drug will be improved by CD complexation, because it will be dissolved faster, will attain a higher solubility, will be absorbed faster [the time between administration and onset of biological effect (e.g., pain reduction) will be shorter], and will result in a more complete absorption.

• Its melting point temperature is below 250  C (otherwise the cohesive forces between the molecules are too strong). • The molecule consists of less than five condensed rings. • Its molecular weight should be between 100 and 400 (with smaller molecules the drug content of the complex is too low; large molecules do not fit the CD cavity). Strongly hydrophilic, too small or too large molecules (e.g., peptides, proteins, enzymes, sugars, polysaccharides, etc.) generally cannot be complexed. Nevertheless, when large water soluble molecules contain appropriate complex forming side-chains (e.g., an aromatic aminoacid in a polypeptide) it will react with CDs in aqueous solutions, resulting in modified solubility and stability (e.g., the stability of an aqueous solution of insulin or many other peptides, proteins, hormones, and enzymes is significantly improved in presence of an appropriate CD). An unavoidable limiting factor in selecting the drug for complexation is the dose of the complex that has to be administered. A fundamental requirement is that the mass of a tablet should not exceed 500 mg. Since the drugs to be complexed have molecular weights between 100 and 400, and the CDs have rather large molecular weights (972, 1132, and 1297 for -, -, and CDs) a 100 mg complex contains only about 5–25 mg of active ingredient. If the single dose of a drug is not more than 25 mg then even a complex of 5% active substance content can carry the necessary dose in a single tablet of 500 mg weight. Otherwise the possibility of a powder sachet or sparkling-tablet formulation has to be taken into consideration. Thus, in the case of complex forming drugs, the relationship of the required dose and the molecular weight determines the feasibility of oral administration in CD complexed form. Similarly, the volume of an injection should be less than 5 ml, or even better not more than 2–3 ml (i.e., to dissolve the necessary amount of the drug in 2–3 ml of 40% HPBCD solution 800–1200 mg HPBCD can be used). In liquid formulations the use of CD derivatives in large excess is possible. For example, in the case of prostavasin injection the molar ratio of prostaglandin E1 to CD is 1:11 (20 g PGE1 + 646 g CD/dose). In slow infusion the parenteral administration of several grams of a noncrystallisable CD derivative is possible. For example, in a liquid itraconazole formulation the weight ratio of the drug to hydroxypropyl CD is 1:40! One single intravenous infusion dose of the commercial product Sporanox contains 250 mg itraconazole and 10,000 mg HPCD. A 3000 I.U. D3 -vitamin tablet contains only 0.075 mg cholecalciferol. A prostarmon-E tablet contains only 0.5 mg PGE2 . The active ingredient content of a nitroglycerin tablet is 0.5–4 mg. These and similar drugs are ideal for CD complexation, but even the 20 mg piroxicam containing Brexin tablet is a widely marketed, successful product. If the Ka stability constant of a complex is low (less than 102 mol−1  the existence of the complex can be evidenced in solution, but removing the water the obtained product is not seldom only a mixture (e.g., a coprecipitate) which contains the host and guest in an intimately fine dispersion. Removing the water is also an important component, as the driving force for complexation is

Cyclodextrins and Molecular Encapsulation

294 eliminated: the repulsive forces between water and the hydrophobic drug. Upon contacting with water the complex formation is an instantaneous process (i.e., in solution the guest is really included in the CD cavity, and the dissociation–association equilibrium is reached within seconds). In such cases the guest is not protected against external destructing factors, like oxygen or humidity, but if the guest is stable enough, only its low solubility means problems. Such intimate mixtures can be utilized for preparation (e.g., solid formulations of improved bioavailability). If, however, the guest is instable then only a full, also in anhydrous state prevailing complexation can help. In the case of extremely high complex stability constants (over about 104 M−1  the bioavailability can even be reduced. The complex is practically not absorbed, only the released, molecularly dispersed (=dissolved) drug molecules. In such cases the co-administration of an even better complex forming competitor molecule (e.g., phenylalanine) can help. Hundreds of published examples illustrate the stabilizing, solubility, and bioavailability enhancing, side-effect reducing, and advantageous technological effects of CD complexation of the instable, poorly soluble, locally irritating drugs. Table 2 illustrates a selection of CD containing drugs, approved and marketed in various countries.

5.2. Food Industry Flavor substances are generally volatile, easily deteriorating substances. Most of them (e.g., terpenoids, phenylpropan derivatives) form stable complexes with CDs and in dry complexed form remain stable for long time, without any further protection, at room temperature. Such powder flavors are approved, produced, and used in several countries

like France, Japan, and Hungary. For example lemon-peel oil/CD complex mixed with powdered sugar is used in pastries or spice-flavor mixtures complexed with CDs in preparation of canned meat and sausages. Peppermint oil/CD complex is used in chewing gum, etc. In Germany garlic oil/CD complex is marketed as odorless dragées (to substitute garlic and a number of various unstable garlic preparations, consumed to reduce blood cholesterol level). CD cannot be considered tasteless. Its taste threshold value is lower than that of sucrose (detection: 0.039% cf. 0.27%; recognition: 0.11% cf. 0.52%). An aqueous solution of a 0.5% CD was as sweet as sucrose, and a 2.5% solution was as sweet as a 1.71% sucrose solution. When CD is used in food processing, its sweetness cannot be ignored. Sucrose sweetness and CD sweetness are additives. Emulsion stability, water retention, and storability can be improved in many cases by addition of cyclodextrins to the food component or food. A solid composition containing sugars and amino acids for food or pharmaceutical purposes is known to undergo browning and caking. This can be prevented by incorporating in such a composition at least 40% of an oligo- or polysaccharide, such as dextrin, starch, or CD, with a water content of not more than 3%. For example the addition of 20% CD to a powdered juice consisting of anhydrous glucose, sodium l-aspartate, dl-alanine, citric acid, and inorganic salts resulted in a product of excellent stability. After 30 days at 40  C there was no apparent discoloration or caking. The control, without CD, began to cake on the second day and to turn brown on the fourth. CD can be utilized for the preparation of stable water-inoil emulsions, such as mayonnaise and salad dressing. Natural food coloring components in tomato ketchup can be stabilized by adding 0.2% -CD. The ketchup thus prepared

Table 2. Some approved and marketed CD-containing pharmaceutical products. Drug PGE1 -CD 20 g/amp. PGE1 -CD 500 g/amp. PGE1 -CD OP-1206-CD Piroxicam-CD Garlic oil-CD

Trade name

Formulation

Indication chronic arterial occlusive disease, etc. controlled hypotension during surgery induction of labor Buerger’s disease Analgesic, antiinflammatory

Company/country

Prostavasin

intraaterial

Ono, J. Schwarz, D.

Prostadin 500

infusion

Prostarmon E Opalmon Brexin, Cicladol Xund, Tegra, Allidex, Garlessence Ulgut, Lonmiel Mena-Gargle Glymesason

sublingual tablet tablet tablet, sachet, and suppository dragées

antiatherosclerotic

capsules garling ointment

anti-ulcerant throat disinfectant analgesic, antiinfammatory

Ono, J. Ono, J. Chiesi, I, Masterpharma, I.D., B., NL., etc. Bipharm, Hermes, D., Pharmafontana, H., D., USA Teikoku, J., Shionogi, J. Kyushin, J. Fujinaga, J.

Ono, J.

Benexate-CD Iodine-CD Dexamethanose, Glyteer-CD Niroglycerin-CD Cefotiam-hexatil-CD New oral cephalosporin (ME 107)-CD Thiaprofenic acid-CD Chlordiazepoxide-CD Hydrocortisone-HPCD

Nitropen Pansporin T Meiact

sublingual tablet tablet tablet

coronary dilator antibiotic antibiotic

Nippon Kayaku, J. Takeda, J. Meiji Seika, J.

Surgamyl Transillium Dexacort

tablet tablet liquid

Roussel-Maestrelli, I. Gabor Ar. Icelandic Pharm., Isl.

Itraconazol-HPCD Diclofenac-Na

Sporanox Voltaren Ophtha

liquid eyedrop

analgesic tranquilizer mouthwash against ahpta, gingivitis, etc. AIDS, oesophagal candidiosis anti-inflammatory

Janssen, B Novartis CH,D,F, etc.

Cyclodextrins and Molecular Encapsulation

did not discolor on heating at 100  C for 2 hours, whereas the control did. Addition of CDs to emulsified foods or cheese can increase water retention and shelf life. In processed meat products CD improves water retention and texture. The addition of 0.005 to 1% CD to canned citrus products prevents precipitate formation, which is mainly due to the poorly soluble, bitter tasting hesperidin and naringin. When the concentration of these substances is too high in grapefruit juice resulting in a bitter taste, they can be selectively removed by passing the fruit juice through a cyclodextrin polymer filled chromatographic column. In Belgium low-cholesterol butter is produced. The molten butter is mixed with CD which does not react with triglicerides but forms complexes with cholesterol, and the CD complex is easily removable from the butter. More than 90% of the cholesterol can be removed in one step. The butter does not retain any CD. Other low-cholesterol milk products, like cheese, cream, or even low-cholesterol eggs, are produced by this technology. CDs, incorporated into food packaging plastic films very effectively, reduce the loss of aroma substances. Incorporating fungicide CD complexes into films—for example, packaging of hard cheeses—significantly elongates the shelf life of the product by inhibiting the rapid development of mold colonies on the surface of the packaged cheese.

5.3. Cosmetics and Toiletry Industry Many active principles of interest in cosmetics have been complexed with cyclodextrins achieving very positive, interesting results. For instance, the inclusion of retinol in hydroxypropyl--cyclodextrin leads to a water-soluble product, sufficiently stable, of higher bioavailability and lower toxicity compared to free retinol. Similarly to retinol, it is possible to include in cyclodextrins almost all (fat-soluble) vitamins, with unquestionable advantages as concerns their stability and bioavailability. The main advantages of using CD complexes in functional dermocosmetics are similar to those in the pharmaceutical formulations: • increase of the bioavailability of incorporated active components (vitamins, hormones, glycolic acid, depigmenting agents, etc.) • protection from oxidation (vitamin A, retinol, hydroquinone, arbutin, DHA, etc.) • reduction of toxicity and aggressivity (glycolic acid, hydroquinone, essential oils, fragrances, etc.) • formation of hydrosoluble complexes even if the incorporated molecule is liposoluble • increased stability of emulsion and gel products • easy formulation (the complex is added to the final formulation, dispersing it in the aqueous phase at 35– 40  C) • no added preservatives Some special advantages were observed for the sunscreen agent/CD complexes: • increased stability (photostability) of cream formulations • decreased cytotoxicity

295 • increased/improved functional activity • reduced smell • reduced fabric staining Thus the markedly increased photostability is associated with decreased phototoxicity and/or photosensitization risk coming from the production of free radicals primed by the sun photochemical attack on the filter molecule itself. Cyclodextrins are suitable not only for real skin treatment products, but also for the formulation of make-up cosmetics which are to stay on the face and skin for a number of hours (about one whole day). CD complexation can be used to stabilize emulsion, color, and perfume or aroma. For instance, most perfume concentrates, such as rose oil, citral, and citronellal, can form complexes and can be used in any solid preparation, such as powdered detergents. The collateral irritant effect caused by a scent in a shampoo can be reduced through the use of CD complexes. A solution of iodine-CD can be used as a deodorant for the body, for baths, or as a refresher for the oral cavity. CD complexes with detergent molecules can act as antifoam agents, etc.

5.4. Textile Industry The first cellulose-CD copolymer was described in 1980. Alkali-swollen cellulose fibers were reacted with CD and epichlorohydrin. The chemically bound CD retained its complex forming ability and could be “loaded” with biologically active guests, like drugs (to prepare medical bandages), insect repellents, antimicrobial agents, etc. The drug is adsorbed partly to cellulose and partly complexed with CD. Upon contact with skin the fabrics made from such fibers ensure immediate drug release from cellulose and a sustained drug release from CD. -, -, and CD has been grafted onto a nonwoven polypropylene support. The fabric was first activated by electron beam irradiation, followed by the graft polymerization of glycidyl methacrylate, a polymerizable monomer carrying epoxide groups. CDs were then fixed onto the fibers by reaction with the epoxide groups. The final CD content after this second step of the reaction increased to 150 mmol per gram of support (20% in weight). 2-Naphtol and p-nitrophenol were sorbed in batch and dynamic filtration systems by the CD grafted supports. Polycarboxylic acids such as citric acid, 1,2,3,4-butanetetracaboxylic acid, or polyacrylic acid have been used as cross-linking agents of cyclodextrins depending on the reaction conditions, and according to the adapted experimental process, polyesters of CDs or textiles carrying CDs could be obtained. Permanent fixation of CDs onto textiles occurred either through reaction of the functional groups of the fibers (covalent bonding), or through the formation of a crosslinked polymer that is tangled up in the fibers (noncovalent bonding) (Fig. 11). Presently the use of monochlorotriazinyl-CD seems to be the most promising approach. CD was condensed with cyanuric chloride in an aqueous medium at 0–5  C in the presence of NaOH to give a 4-chloro-6-hydroxy-triazin-2yl-CD Na salt with degree of substitution of active Cl 0.4. This product is a reactive CD derivative that can be

Cyclodextrins and Molecular Encapsulation

296 a

O

O

N N

N

ONa

triazinyl b O C

COOH

O

C H

C

H C

O

O

COOH

1,2,3,4-butanetetra-carboxylic acid

the CD is free again to trap odors. CD textiles can be recharged selectively with the most diverse guest substances by dipping or spraying. Particularly interesting is the treatment with fragrances and antimicrobial or pharmaceutical preparations which can be released continuously over a long period on exposure to moisture. The higher the level of moisture, the more guest substance is liberated. CDmodified fabrics, therefore, “respond” to varied ambient conditions (Fig. 12). In Germany 5 million people suffer from neurodermitis, 2 million from psoriasis, and a further 3 million from dermal diseases of occupational origin (like hairdressers’ allergic reactions on the hands, athlete’s foot, etc.). The corresponding article of clothing (underwear, gloves, socks) can be loaded with the necessary antifungal, antihistaminic, antiinflammatory, etc. agents. A further application is towels, which start smelling after drying one’s hands. Also bed linen may release a pleasant odor after going to bed. Curtains with cyclodextrins

c H C

O CH2

OH

O CH2

+ glyceryl-ether

perfume

Figure 11. Fiber-cyclodextrin binding types: (a) triazinyl chloride reacted with cyclodextrin and the product reacted with the fiber; (b) polycarboxylic acid reacted with the fiber, the formed ester converted to cyclic anhydride, and then it reacted with cyclodextrin; and (c) the cyclodextrin and epichlorohydrine reacted with strong alkaly treated cellulose.

covalently fixed to nucleophilic substrates by a condensation reaction carried out by the usual finishing procedure. This new type of surface modification means a permanent transfer of CD properties to the treated materials. The finish is applied by standard methods—analogous to those of reactive dyeing—on conventional equipment. The dissolved monochlorotriazinyl-CD (MCT-CD) is drawn up from a dipping bath, the textile is squeezed, and the CD is fixed on the fiber at elevated temperature. Unattached CD is simply removed by washing. Both alkaline and acidic methods for MCT-CD finishes have been developed. While fixing temperatures and times are the main parameters for the fixing reaction, other variables such as dampness of the fabric and the MCT-CD concentration are also important. An MCT-CD finish is wash resistant. Cotton for T-shirts, sports socks, and so forth is particularly amenable to CD modification. However, the latest studies also reveal that very good fixing results may be obtained with blended fabrics, such as cotton/polyurethane and cotton/polyurethane/polyamide. This is especially important for top-of-the range lingerie. Given suitably adapted processes, moreover, wool can also be finished with MCT-CD. Encapsulation of sweat or cigarette smoke greatly reduces the intensity of odors on articles of clothing or furniture textiles. In the washing machine, the encapsulated substances are washed out, and

A

+

COOH

fatty acid

B

COOH

+ free perfume

Figure 12. Guest exchange, for example, perfume release and sweat component entrapment. (A) Loading the fixed cyclodextrin with perfume. (B) During wearing, upon the effect of sweat, humidity, and excreted short chain fatty acids, the guest will be exchanged, the perfume gets released, and the fatty acids are entrapped.

Cyclodextrins and Molecular Encapsulation

can be used to improve the air in rooms or offices with smokers. The identification of the organic compounds excreted from patients through their skin enables new methods in medical diagnostics. Up to now blood or urine tests have normally been used due to the problems of taking a sweat probe from a patient. This becomes easy by wearing for 24 hours a T-shirt which contains chemically bound CD. After extraction of the T-shirt, the substances— characteristic, for example, of metabolic disorders—excreted through the skin are detectable by gas chromatography. The complexation of organic substances from sweat results in a preconcentration and their identification becomes easier. Mixing CD or CD–polymer fragrance complexes to the melt mixtures of synthetic fiber polymers (e.g., polyester) and weaving the fabric from such fibers, washfast fragrant fabrics can be produced.

5.5. Biotechnology The majority of biotechnology processes mean an enzymecatalyzed transformation of a substrate in aqueous medium. The main difficulties which used to arise were the following: • The substrate is hydrophobic, sparingly (or hardly) soluble in water. • The enzyme or the enzyme-producing microbial cells are sensitive to the toxic effects of the substrate or to inhibitors which can even be the product of the transformation. • The substrate or the product is unstable under the conditions of the enzymatic transformation. • Isolation of the product from the very heterogeneous system is difficult. Cyclodextrins and their derivatives enhance the solubility of complexed substrates in aqueous media and reduce their toxicity, but they do not damage the microbial cells or the enzymes. As a result, the enzymatic conversion of lipophilic substrates can be intensified (accelerated, or performed at higher substrate concentrations), both in industrial processes and in diagnostic reagents, the yield of productinhibited fermentation can be improved, organic toxic compounds are tolerated and metabolized by microbial cells at higher concentrations, and compounds in small amounts can be isolated simply and economically from complicated mixtures. Examples illustrate the rapidly growing and promising uses of cyclodextrins in various operations: the intensification of the conversion of hydrocortisone to prednisolone, the improvement in the yield of fermentation of lankacidine and podophyllotoxin, the stereoselective reduction of benzaldehyde to l-phenylacetyl carbinol, and the reduction in toxicity of vanillin to yeast, or organic toxic substances to detoxificating microorganisms. In the presence of an appropriate cyclodextrin derivative (e.g., 2,6-dimethyl-cyclodextrin), lipidlike inhibitor substances are complexed. The propagation of Bordatella pertussis and the production of pertussis toxin therefore increases up to a hundredfold. Cyclodextrin and their fatty acid complexes can substitute mammalian serum in tissue cultures.

297 In recent years Leprae bacillus (Mycobacterium leprae) was considered to be not cultivable under in vitro conditions. The most important energy source for bacillus is palmitic (or stearic) acid which, however, cannot penetrate through the thick strongly hydrophilic shell of the Mycobacterium. Solubilizing, however, the fatty acids (or fatty alcohols) with dimethyl--cyclodextrin, the Mycobacterium can be cultivated in vitro, on synthetic medium. This discovery will facilitate the screening of drugs against similar difficultly cultivable microorganisms. The biological remediation of carcinogenic polyaromatic hydrocarbon polluted soils is a very slow process, because the dangerous polluting hydrocarbons are adhered very strongly to the soil particles. They are not available for the soil micro-organisms. Adding CDs (for example, the randomly methlyated CD) to such soils, the CD mobilizes the hydrocarbons and makes them available for the micro-organisms, resulting in an accelerated remediation of the soil.

5.6. Chemical Industry Fragrant paper or paper containing protective substances can be prepared using CD complexes of perfumes, insecticides, rust inhibitors, mold- and mildew-proofing agents, fungicides, and bactericides. For example, fenitrothion CD complex sprayed on a wet paper web, passed between drying rollers heated to 100  C, resulting in an insecticide containing paper, was shown to be effective for more than 6 months. Conservation of wood-made products which are prone to attack of micro-organisms, like frames of windows, doors, and buildings constructed from wood, usually has been made by impregnating the wood with fungicides, which are generally water insoluble. Therefore they had to be dissolved in organic solvents. After impregnation the solvents escaped into the atmosphere. A new process dissolves the fungicides in aqueous CD solutions; the impregnation and conservation of the woods can be made without using organic solvents. Emulsion-type coatings (paints) contain emulsion polymer binders, to give—after drying—a resistant, continuous protecting film on the coated surface. To ensure the formation of a good film, the applied layer has to contain various compatible components, like solvent, pigment, thickener, and binder. The rheological properties of the paint are determined by the thickener, which is usually a hydrophobically modified polymer, like polyurethane, polyacrylamides, cellulose ethers, etc. To avoid a concomitant too high viscosity (which makes difficult the formation of a uniform coating of the surface to be covered) viscosity suppressors has to be added to the emulsion. Adding organic solvent to such emulsions, the viscosity can be reduced, but use of organic solvents has to be avoided because of safety, health-damaging, environmental polluting effects. Also surfactants can strongly reduce the viscosity of such emulsions, but their use results in the formation of a less resistant coating. The viscosity enhancing effect of hydrophobically modified macromolecules in aqueous emulsions is based on the hydrophobic–hydrophobic interactions between these molecules. Adding CDs to this emulsion the CD molecules

298 will associate with the hydrophobic sites and, being strongly hydrated, inhibit the association of the macromolecules, resulting in a strong reduction of the viscosity. The RAMEB was shown to be the most effective viscosity suppressor. The CD complexes are compatible with thermoplastic resins. Mixing a dry pulverized CD complex of a perfume, for example, a geraniol CD complex, with a thermoplastic resin (polyethylene), and molding it yielded plastic products with a long lasting (at least 6 months) fragrance. Rapid loss of the perfume by volatility and thermal decomposition can be avoided in this way. Mixing CD complexes of thymol, eugeneol, isobutylquinoline, etc., to molten PVC (=polyvinylchloride) a natural leather odor emitting (leatherlike) material was prepared (e.g., for automobile door internal coverings). The epoxyresin adhesives are produced and stored as separated components mixed just before utilization. Complexing the curing agent (polymerization catalyst) with CD a one-package composition can be prepared. Upon layering such a composition between metal plates and heating to 130  C for 5 min, binding will take place. Properties of cyanacrylate adhesives can be improved significantly by heptakis(2,6-di-O-butyl-3-O-acetyl)-CD. The ethyl-2-cyanacrylate monomer is stabilized with 20 ppm phosphoric acid, 20 ppm SO2 , and 100 ppm hydrochinon. Various amounts of the dibutyl-acetyl-CD were added to this mixture, and using an adhesive, for example, to bind hard cartoon papers by quick heating to 200  C, the polymerization (binding) of cyanacrylate has been accelerated, and the tear-strength of the binding increased significantly. Explosive substances, like nitroglycerin or isosorbidedinitrate, when complexed can be handled without any danger. The pernitro-CD complex of the extremely explosive nitramine can be used as a controlled missile propellant. Important properties such as relative sensitivity and the fog of silver halide containing photographic materials can be improved by adding CDs to the light sensitive photographic gelatin layers. Additives, dyes, stabilizers, and fog inhibitors used in the photographic industry should be fixed to a certain layer of the film or photopaper. This can be achieved by using derivatives with “heavy” side chains. It seems to be more convenient to prepare the water soluble polymer complexes of these substances. In complexed form their mobility is markedly reduced, and they become fixed to the required layer. A diminished diffusion can be observed either on preparing, processing, or storing the film. Another advantage of the soluble polymer complex is that poorly soluble or even water insoluble stabilizers can be applied to the film in aqueous solutions. A CD-gelatin composition as photographic layer shows a lower water absorption and accelerated diffusion of the developing agents. Silane resins are made using a 1,5-cyclooctadieneplatinum catalyst. Once the catalyst is added, the reaction proceeds. Complexing the catalyst with CD, it became possible to produce a single component heat-curable silicone with a shelf life exceeding 7 months since complexation prevented immediate catalysis and reaction. No gelation occurred at ambient temperature, only when the mixture was heated up to 150  C.

Cyclodextrins and Molecular Encapsulation

Active ingredients of pesticides (insecticides, fungicides, herbicides, etc.) can be complexed with the same technologies and same consequences as drug molecules. The selectivity of an insecticide can be improved. For example, complexing an insecticide, which kills herbivorous insects, will not be toxic for the honey-bee. Stability against sunlight accelerated decomposition of pyrethroids can be improved ensuring a longer lasting insecticidal effect. Wettability or solubility of poorly soluble benzimidazole type antifungicides can be improved, allowing the reduction of the dose and the pollution of the environment. The possibilities of utilization of CDs in pesticide formulations have already been elucidated in numerous scientific papers. However, until recent years the price of CD did not allow its utilization for such purposes because the pesticide industry is very cost sensitive. By the end of the 20th century the price of CD dropped to such a low level that the price will no longer be a restricting factor in its utilization for pesticide formulations. The preparative- and technical-scale chiral resolution of enantiomers is one of the biggest challenges in industrial utilization of CDs. The utilization of CDs in this area is still in its infancy but some results are encouraging. For example, a continuous method for separation of dichlorophenyldichloronapthalenone enantiomers has been published using different CD derivatives in a multiple solvent extraction technology. A continuous, easy to scale up preparative chiral separation device, a slowly moving “chiral” belt coated with 3,5-dimethyl-phenyl-carbamoyl-CD, enters into and passes through different solutions in separate troughs. The solutions contain different solvents: trough 1 contains the racemic analyte (RS-oxprenolol), troughs 2 and 3 are filled with a hexane:i-propanol mixture for enantioselective desorption of analyte, while only hexane is in trough 4 for rinsing/regenerating the belt. The continuous process results in 2 hours in the enrichment of R-oxprenolol isomer in trough 1, while trough 3 becomes rich in the S-isomer (about 68%). Despite the encouraging results in the preparative scale resolution a reliable and economic technology for the routine enantiomer separations in 1 or 10 g pure enantiomer per batch still remains a challenging task of the future.

5.7. Analytical Chemistry 5.7.1. Chromatographic Separations CDs and mainly specific derivatives are widely used in chromatographic separations. In 1990–2000 about 25 papers were published monthly on successful application of CD in gas or liquid chromatography. In gas chromatography the CDs are used only in the stationary phase, while in high pressure liquid chromatography (HPLC) the CD is used either as dissolved in the mobile phase or bound to the surface of the stationary phase. In capillary zone electrophoresis the most used chiral selectors are the CDs and their anionic, cationic, or alkylated derivatives.

5.7.2. Diagnostics The actually used and marketed chiral separation products include HPLC, gas chromatographic columns, and capillary electrophoresis buffer additives.

Cyclodextrins and Molecular Encapsulation

In medicinal diagnostic kits frequent problems are the limited, uncertain solubility of some essential component, the stability, and the storability of the kit. Complexing with CDs as the typical components, such problems can be eliminated. Frequently intensity of color reactions or of fluorescence can be significantly enhanced by the presence of CDs, resulting in a higher sensitivity of the method. In the following a few examples illustrate the application of CDs in clinical sample preparation, in enhancement of sensitivity of analytical methods, CDs used for solubilization and/or stabilization of reagents, and CD based sensors and indicators. In clinical chemistry: • CD is used for the selective and highly effective extraction of serum Digoxine in one single run enabling the convenient determination of the drug in biological samples by the RIA method. • CD is also used for the determination of steroid hormones in human serum traditional solvent extraction of serum samples, followed by a re-extraction of the steroids into an aqueous cyclodextrin solution. • A pre-column containing immobilized CD-sulfate was successfully applied for the on-line enrichment and separation of heparin binding proteins. • The treatment of lipemic serum with aqueous solutions of CDs will produce an immediate and quantitative precipitation of the lipoproteins. CD was found to precipitate chylomicrons, very low, low, and high density lipoproteins. • CDs can selectively remove disturbing components from complex biological samples. The presence of fatty acids in blood samples disturbs colorimetric analysis and causes an underestimation of the actual Ca2+ level. CD was found to eliminate this interference, completely. • An automated LDL-cholesterol assay has been recently marketed that employs CD together with Mg ions to effectively mask interfering lipoproteins present in clinical samples. Through complexation of analyte and/or the colorforming reagent by CDs in the spectrophotometric determination of a wide variety of compounds, the sensitivily of the method is significantly improved. For example: • The spectrophotometric determination of copper in leaves and in human hair was improved by CDs, as the sensitivity of the color reaction of Cu(II) and mesotetrakis-(4-methoxy-3-sulfophenyl) porphyrin was enhanced by 50% in the presence of CD. • In the determination of microamounts of Zn based on the Zn-dithizone color reaction sensitized with CD, the apparent molar absorptivity at 538 nm is 8.37 times larger than that in the absence of CD. • CDs often can result in significant enhancement of the fluorescence or phosphorescence of the complexed analyte. A substantial enhancement of the fluorescence emission of aflatoxins in the presence of aqueous solutions of - and -di-O-methyl-CD and hydroxypropylCD was observed, which enabled the determination of minute amounts of these toxins in food samples.

299 • Determination of selenium in biological matrices was improved using CD by which the relative fluorescence of the reagent 4,5-benzopiaselenol is about 150-fold higher than that without CD. • The application of CD in beryllium analysis allowed the determination of 10–40 ng/mL beryllium, compared to 60–500 ng/mL in the absence of the CD. • CD-assisted spectrofluorimetric assay enabled the detection of mescaline in the 0.8–1.4 ppm range. • The application of CD enabled a limit of detection of 0.07 g/ml zirconium. • CD stabilizes dichlorofluorescein against autoxidation and enhances its fluorescence, a dual benefit that can be used to improve the determination of serum uric acid and glucose. • Improved sensitivity was reported on trace cadmium determination from hair, nails, water, etc., which was based on the enhancement of the chromogenic reaction due to CD complexation. This method enabled the determination of about 0.4 ng/ml cadmium. The solubilizing effect of cyclodextrin complexation for lipophilic reagents or analytes provides the possibility to replace organic solvents and detergents with aqueous CD solutions. For example: • Complexation of the unstable, insoluble substrate l--glutamyl-4-nitroanilide with CDs eliminated the errors occurring in the determination of -glutamyltransferase activity caused by its poor solubility. • Poorly soluble antigens can be solubilized by CD in immunoassay used for determination of progesterone in human serum by the solid phase EIA (enzyme immunoassay) method. • Fluorescent labeling of proteins done with dansylchloride organic water-soluble complex with CD enables the labeling of proteins in aqueous solutions. • SIN-1A/CD complex is a stable analytical standard suitable for reliable determination of NO and calibration of NO detectors. • By complexing highly unstable ortho-phenylenediamine, an immunodiagnostic reagent with CD, and simply tabletting the solid complex, a stable reagent with extended shelf life was prepared.

5.7.3. Sensors CD-based supramolecular assemblies or covalent new chemical entities have been developed as selective sensors, detectors, and indicators. For example: • A benzene-vapor detecting piezoelectric quartz crystal was constructed by coating the surface with 2,6O-tert-butyl-dimethyl-silyl-CD. This detector is highly selective for benzene even in large excess of methane, propane, butane, pentane, ammonia, nitrobenzene, and toluene. • CD fixed on a quartz oscillator surface was shown a very sensitive sensor for cholesterol suitable for determination of blood cholesterol level. • Various compounds including steroids, bile acids, terpenes, etc. have been selectively detected with a specifically designed cromphore bearing CD derivatives.

Cyclodextrins and Molecular Encapsulation

300 • Voltammetric responsive sensors based on organized self-assembled CD derivatives in monolayers on gold electrodes were studied for analysis of electroinactive organic analytes. • A fiber-optic CD-based sensor was developed using laser excitation and fluorescence detection, with CD as the reagent phase immobilized at the tip of an optical fiber. The sensitivity of the fiber-optic CD-based sensor was found to be 14 times greater than that of a bare optical fiber. • Lipophilic cyclodextrins were used for construction of guanidinium ions and choline, acetylcholine selective sensors.

+

N

O

N

N

N+

O

N+

+

N III

O2N

NO2

NO2

O2N +

+

+

+

Trans

5.7.4. Standards CD complexation offers the possibility for preparation of flavor standards. There are cases when human taste and smell sensors cannot be substituted even by the most sophisticated analytical instruments. In organoleptic evaluation, comparison of identical food products of different origin—for example, beers— the dosing of extremely small but exactly known amounts of flavor components (for example ppm amounts of mercaptanes, butyric acid, hydrogen sulphide, etc.) is necessary. To store and dose these volatile, unstable substances becomes very simple when they are complexed with CDs, diluted with inert excipient to the appropriate grade, and filled in known amounts into hard gelatine capsules. Adding the content of one or more capsules of different flavors to the sample at sensoric “titration,” the differences between the products can be expressed unambiguously and numerically.

6. CYCLODEXTRIN BASED NANOSTRUCTURES Cyclodextrins belong to the most appropriate rotaxane forming molecules. A long slim guest molecule can be threaded through the CD cavity. Then both ends can be terminated by bulky groups or the terminal group can be ionized and therefore the threaded molecule cannot slip out of the cavity (Fig. 13). Upon various environmental effects (pH, irradiation, electric field, etc.) this threaded molecule may rotate O2N NO2

NH2

NH

O2N O 2N +

SO3Na H 2O

+

NO2 O2N

H2N HN

NO2

O2N NO2

Figure 13. Principle of the CD-rotaxane formation.

λ = 430 nm

+

λ = 360 nm

+

+

+

Cis

Figure 14. A light driven shuttle.

around its axis; otherwise its mobility is restricted. Similarly the CD-ring mobility is also restricted; it can move only along the axis. This way a molecular “shuttle” or “switch” can be constructed. Threading an azobenzene derivate through an CD ring, the azo-bridge will be allocated in the CD cavity. Attaching 2,4-dinitrophenyl-bipyridinium “stoppers” to both ends of the threaded axis molecule, a shuttle rotaxane is formed. Irradiation of its aqueous solution with UV light (360 nm) causes isomerization of the azobenzene moiety from a trans to cis configuration: the isomerization can be reversed when the solution is irradiated with visible light (430 nm). This configuration change is reflected in a dramatic alteration of the geometry (Fig. 14). The CD ring resides around the azo-bridge; when the azobenzene unit adopts the trans configuration the CD ring migrates toward one of the bis-(methylene) spacers. The photoswitching process is reversible, and this is an excellent example of a light driven molecular shuttle. Threading a long slim guest through a number of CD rings, a “molecular necklace” can be prepared (Fig. 15). Recently the nylon/CD complex, by reacting the CD complex of hexamethylene diamine with CD-complexed diacylchloride, has been published. This way new materials with quite interesting properties can be produced. “Molecular tubes” can be prepared for example by complexing polyethylene glycol-bis-amin with CD; then the formed polyrotaxane is reacted with 2,4-dinitrofluoro benzene. This way both ends of the long chain guest are terminated by bulky groups. Reacting this polyrotaxane with epichlorohydrine the vicinal CD rings will be interconnected through glyceryl bridges between the primary and secondary sides of the CDs. Finally, upon the effect of strong alkali, the dinitrofluorobenzene groups will split off and the long polymer chain will slip out from the polymeric tube (Fig. 15).

Cyclodextrins and Molecular Encapsulation

H2N

O

O

O

O

O

O

O

O

301 O

NH2

Polyethyleneglycolbisamine (MW > 2000) αCD

H2 N

O

O

O

O

O

O

Pseudopolyrotaxane or Molecular necklace F O2N O2N

NH NO2

O

O

O

O

OH

O2N

NH NO2

O

OH

O

O

O

O

NH2

NO2

O

O

O

O

HN O2N

NO2

HN O2N

NO2

H2C CHCH2Cl / 10% NaOH O OH OH OH

Polyrotaxane or Molecular necklace OH

O

O

O

OH

Threaded Molecular tube

O

OH

O

O

O

OH

O

OH

25% NaOH

OH

OH

OH

OH

OH

OH

OH

OH

OH

OH

Molecular tube

Figure 15. Scheme for the sequential formation of CD pseudopolyrotaxane, polyrotaxane (molecular necklace), and molecular tube.

If the threaded molecule itself is closed to form a closed ring, then the formed structure is called catenane or, if it includes more than one CD ring, polycatenane (Fig. 16). Metal ions can be complexed with CDs in different ways: • The metal ion reacts with the hydroxyl groups of the CD molecule. • The metal ion forms a coordination complex with the usual organic ligands and this coordination complex will be included in the CD cavity. • The metal atom is bound covalently in a metallorganic compound which will form a regular inclusion complex with a CD molecule. • The metal ion is bound by an anhydro-CD. In the first case, a hydroxo complex is not an inclusion complex; more likely it is an outer sphere complex (e.g., Cu2+ or Mn2+ ions in alkaline solutions form such CDhydroxo-metal complexes). The second case means the formation of ternary complexes: CD + organic ligand + metal ion. This is a real second sphere coordination metal complex. For example, a ferrocene is a coordination complex which consists of an iron

ion, sandwiched between two cyclopentadiene molecules, forming a stable iron-coordination complex. This coordination complex can form a true inclusion complex with CDs and this inclusion strongly modifies the physical and chemical properties of the included coordination complex. The technecium-99 ethylcysteinate dimer for brain imaging or the technecium-99 nitrido-dithiocarbamate for myocardial scintigraphy can be used only after solubilizing them with an appropriate CD. In the third case, the complexation of organometallic compounds—this is a binary complex—also results in the modification of important properties of the included compound (e.g. in pharmaceutical preparations). The ferrocene complexes can be prepared easily, in crystalline form, with good yield. The excess of sublimable ferrocene can easily be removed by vacuum sublimation, while the CD-bound ferrocene is stable up to the temperature of the caramelization of CDs. The orientation of an ionizable ferrocene within the CD cavity depends on the pH of the solution (Fig. 17). On the basis of circular-dichroism spectra the ferrocene carboxylic acid was assumed to orient itself inside the CD cavity parallel to its axis, while the ionized carboxylate ion was perpendicular to it (at pH 9 in water). In aqueous solution the antitumor carboplatin forms an 1:1 complex only with CD, but not with - or CD (Fig. 18). The cyclobutane ring penetrates the CD cavity, with additional stability arising from hydrogen bonds between the ammine ligands and hydroxyls. Also dimethylCD forms a similar complex with carboplatin. In contrast, the platinum phosphine complex trans-Pt(PMe3  Cl2 (NH3  forms an 1:1 complex with CD but not with CD, and the hydrophobic trimethylphosphine ligand resides in the CD cavity. The 3,6-peranhydro-2-O-methyl-CD showed a surprisingly high affinity for lead (12 × 107 M−1 ) and for strontium (18 × 104 M−1 ). Eventually this and similar CD derivatives will be utilizable for decontamination or detoxication by sequestering selectively certain metal ions from the contaminated environment, or even from intoxicated organisms. The essence of the photodynamic tumor therapy is that such compounds have to be delivered to the tumor tissues, which upon strong light irradiation become toxic through isomerization, splitting, etc. In this case upon strong light irradiation the photosensitive molecules will become toxic just for the tumor cells. For such targeting of the drug very stable (105 –107 M−1 ) complexes are needed. The duplex or triplex homo- or heterodimers of CDs (constructed only from one or two different CDs) form complexes which are COO-

Fe

[2]-Catenane

[4]-Catenane

Figure 16. Various catenanes.

COOH

Fe

Polycatenane

Figure 17. Orientation of an included ionizable ferrocene depends on the pH of the solution.

Cyclodextrins and Molecular Encapsulation

302 Pt

H3N O

NH3 O

O

O

Figure 18. Structure of carboplatin.

more stable by orders of magnitude than the singular CDs (Fig. 19). Interconnecting two CDs with appropriate bridges such duplex CD derivatives have been prepared which can form stable complexes with photosensitive porfirinoid structures and to transport them to the target organs. In “antennae” bearing CDs receptor specific oligosaccharide units are attached to the CDs, which will be bonded in the living organism only to certain specific receptors. The aim of this effort is to synthetize a receptor-targeting carrier; that is, the drug complexed with an antenna bearing duplex CD would transport the specific drug just to the target organ. CDs are considered to be not very selective receptors for a very large number and variety of relatively small (CD-cavity compatible) molecules. A promising new challenge is the synthesis of such artificial receptors, which will show selectivity for nanometer scale large molecules. One promising approach is molecular imprinting. Polymerizing (with appropriate cross-linking agents) the cholesterol/CD complex and then removing the guest (e.g., by excessive organic solvent extraction) the obtained “empty” polymer will show an interesting specific affinity for cholesterol. No other potential guest will fit so correctly into the cavity of the ordered assembly than cholesterol. A nonimprinted CD polymer hardly bounds any cholesterol. The assembly of CDs is ordered by the template (=guest) molecule, which facilitates the fixing of the position and orientation of the CD units by cross-linking just in the most advantageous position for maximum cooperation in

S

S NH

HN

Cyclodextrine-Trimer

Figure 19. A triplex CD.

binding the guest. Using this principle a large variety of single molecule specific “accommodations” can be memorized by such ordered and fixed assemblies (Fig. 20). The most complicated CD derivatives are synthetized for enzyme-modeling experiments. For all enzymes it is characteristic that they have a substrate binding site and an active, catalytic group. The CDs are similar; the axial cavity is the binding site, and the hydroxyl group can attack, for example, an ester linkage of an included ester-type guest. This enzymelike activity is, however, weak and restricted to only a few well studied examples. Very active and specific artificial enzymes can be constructed by using the CD molecule as a scaffold. The axial cavity can be elongated (e.g., by alkaline substitution on the primary face) and one or more CD hydroxyls can be substituted by active, prosthetic groups. This way artificial hydrolase, isomerase, lyase, transferase, oxydoreductase, etc. enzyme models have been synthesized. In some cases their activity approaches well the activity of natural enzymes. Combining the molecular imprinting technique with enzyme modeling, very probably artificial enzymes of high activity and specificity will be produced, both for industrial applications and for in vivo (e.g., detoxicating) purposes. A CD dimer (Fig. 21) with a linker containing a bipyridil group can form a coordination complex with metal ions. These are very effective at hydrolyzing ester substrates which have two hydrophobic arms that can be fixed by the CD cavities of the catalyst.

+ crosslinking

A

B

C

D

Figure 20. Principle of molecular imprinting: a guest (A) is complexed with a CD in solution (B). Then the CDs will be cross-linked with an appropriate agent (C) forming either a water-soluble oligomer (Mw < ∼10000) or a water-insoluble polymer (Mw > ∼10000). Finally the guest is removed (e.g., by an appropriate organic solvent) from the imprinted structure. The size, shape, and allocation of binding sites are memorized. This cavity hardly fits with other potential guests but shows high selectivity for the original “template” guest.

Cyclodextrins and Molecular Encapsulation

303

S

S

P O

GLOSSARY

O

N N La3+ O O– –O O

O P O

O2N

O

R

R= – R = -CH3

O NO

O 2N

NO2

2

Figure 21. A cyclodextrin dimer which can bind both ends of a substrate and hold it onto a catalytic metal ion (La3+ ) is a very effective catalyst for ester hydrolysis.

Attaching four CD molecules to a porphyrin, or only 2 but in the A–C (or B–D) position, it will bind sufficiently long bifunctional “inclusion” guests (E of F) and keep them stretched with their aliphatic or olefinic section in close proximity to the center of the porphyrin structure. If the nitrogens of prophyrin are coordinated with an Mn(III) atom, then the aliphatic segment of the guest will be hydroxylated or the olefinic segment will be epoxidated (Fig. 22). When the two CDs are linked at the vicinal (A–B, B–C, C–D) positions no guest binding takes place.

A

S

B

NH N S

S

Mn N NH

D

S

C

E HO Mn3+

R

R R

R

PhlO COOH R=

F

O

Mn3+

O R

R PhlO

R

R

O R=

N H CO2H

NO2

Figure 22. Porphyrin carrying four attached CD rings binds strongly to an appropriate two-armed guest and, forcing its oxidizable center near to the Mn3+ ion atom, results in hydroxylation of a –CH2 – unit or epoxidation of an olefinic bond.

Administration routes Mode of incorporation of a drug into the organism: i.v. = intravenous, p.o. = per oral, i.m. = intramuscular, s.c. = subcutaneous, i.p. = intraperitoneal, etc. AUC value (area under curve) The integrated area under the curve blood level versus time (counted from time of administration till end of the observation, e.g. 24 h). Bioavailability The absorption (penetration) of drugs through the absorption sites (gastrointestinal system, skin, lung, mucous membranes, etc.) is frequently far from being complete, and biological utilizability (bioavailability) depends on numerous factors, but the prevailing one is the solubility, and dissolution rate of the administered drug. CTG ase (cyclodextrin trans glucosylase enzyme) This enzyme splits the glucosidic linkages between the sixth, seventh, eight (etc.) glucopyranose units in the amylose or amylopectin molecules (components of starch of any origin) then establishes new glucosidic linkages between both ends of the formed break-down products resulting in formation of cyclic dextrins = cyclodextrins. Cyclodextrin 6,7,8 (or more) glycopyranose units are bound through -1,4-glycosidic linkages. The outer surface of these conical torus-shaped (doughnut shaped) molecules is hydrophilic, in aqueous solution strongly hydrated but the internal cavity of well defined diameter and length is apolar, ready to incorporate appropriate apolar guest molecules. Cyclodextrin catalysis Cyclodextrins similarly to all enzymes have a substrate binding site (the CD cavity) and active catalytic moieties, the secondary and/or primary hydroxyl groups. Substituting one or more CD hydroxyls with appropriate substituents aritificial enzyme models can be synthetized. Cyclodextrin derivatives The hydrogen from one or more hydroxyl group or the complete hydroxyl group are replaced by substituent(s), resulting in ester, ether, deoxy-, amino-, carboxy-, halogen-, etc. derivatives of CDs. Degree of substitution Number of substituents, attached to the CD ring. Guest molecule A molecule, which at least one part of it can penetrate into the CD-cavity, to form an “inclusion” complex. Host molecule A molecule, which has an open cavity of appropriate size and polarity for incorporation of an appropriate “guest” molecule. Inclusion complex A guest molecule (or a part of appropriate size and polarity of it) is included, and bound by noncovalent forces in the cavity of a host molecule. Molecular necklace Rotaxane, which consists of long slim axis guest molecule and several CD-hosts around it. Molecular encapsulation Inclusion complexation. Molecular switch (or shuttle) In a rotaxane the host CD changes its position along the axis guest molecule, upon external effects (light, pH) or the orientation of the (partly) included guest is modified relative to the CD-cavity. Outer-sphere complex The host (cyclodextrin) is linked through a non-covalent bond to a non-incorporated guest (e.g., by hydrogen bond, through the CD-hydroxyls).

304 Rotaxane A long slim guest molecule threaded through one or more CD cavity, then both ends are terminated by “stopper” moieties, which cannot penetrate through the CDcavity, impeding the “slip out” of the axis guest molecule.

REFERENCES 1. D. French, Adv. Carbohydrate Chem. 12, 189 (1957). 2. M. L Bender and M. Komiyama, “Cyclodextrin Chemistry.” Springer-Verlag, Berlin, 1978. 3. “Proc. First Int. Symp. on Cyclodextrins,” Budapest, 1981 (J. Szejtli, Ed.), Reidel, Dordrecht, 1982, p. 544. 4. J. Szejtli, “Cyclodextrins and Their Inclusion Complexes,” p. 296. Akadémiai Kiadó, Budapest, 1982. 5. “Inclusion Compounds” (J. L. Atwood, J. E. D. Davies, and D. D. MacNicol, Eds.), Vols. 1–3, Academic Press, London, 1984. 6. J. L. Atwood, J. E. D. Davies, and T. Osa, Clathrate compounds, molecular inclusion phenomena and cyclodextrins, in “Proc. of Third Int. Symp. on Clathrate Compounds and Molecular Inclusion and the Second Int. Symp. on Cyclodextrins,” Tokyo, 1984, Reidel, Dordrecht, 1985, p. 426. 7. Inclusion phenomena in inorganic, organic and organometallic hosts, in “Proc. Fourth Int. Symp. on Inclusion Phenomena and the Third Int. Sympt. on Cyclodextrins,” Lancaster, 1986 (J. L. Atwood and E. D. Davies, Eds.), p. 455. Reidel, Dordrecht, 1987. 8. “Cyclodextrins and Their Industrial Uses” (D. Duchene, Ed.), Editions de Santé, Paris, 1987, p. 448. 9. J. Szejtli, “Cyclodextrin Technology,” Kluwer Academic, Dordrecht, 1988, p. 450. 10. “Proc. Fourth Int. Symp. on Cyclodextrins,” Munich, 1988, (O. Huber and J. Szejtli, Eds.), p. 576. Kluwer Academic, Dordrecht, 1988.

Cyclodextrins and Molecular Encapsulation 11. “Minutes of the Fifth Int. Symp. on Cyclodextrins,” Paris, 1990 (D. Duchene, Ed.), Editions de Santé, Paris, 1990. p. 722. 12. “New Trends in Cyclodextrins and Derivatives” (D. Duchene, Ed.), Editions de Santé, Paris, 1991, p. 635. 13. “Minutes of the Sixth Int. Symp. on Cyclodextrins,” Chicago, 1992 (A. R. Hedges, Ed.), Editions de Santé, Paris, 1992, p. 693. 14. K. H. Frömming and J. Szejtli, “Cyclodextrins in Pharmacy.” Kluwer Academic, Dordrecht, 1993. 15. “Proc. of the Seventh International Cyclodextrin Symposium,” Tokyo, 1994 (T. Osa, Ed.), p. 554. Publ. Office of Business Center for Academic Societies, Tokyo, Japan. 16. J. Szejtli, Med. Res. Rev. 14, 353 (1994). 17. “Comprehensive Supramolecular Chemistry” (J. Szejtli and T. Osa, Eds.), Vol. 3, p. 693. Pergamon, Oxford, 1996. 18. “Proc. of the Eighth International Cyclodextrin Symposium,” Budapest, 1996 (J. Szejtli and L. Szente, Eds.), p. 685. Kluwer Academic, Dordrecht, 1996. 19. J. Szejtli, Chem. Rev. 5 1743 (1998). 20. S. A. Nepogodiev and J. F. Stoddart, Chem. Rev. 5, 1959 (1998). 21. R. Breslow and S. D. Dong, Chem. Rev. 5, 1997 (1998). 22. K. Uekama, F. Hirayama, and T. Irie, Chem. Rev. 5, 2045 (1998). 23. “Proc. Ninth Int. Symp. on Cyclodextrins,” Santiago de Compostella (J. J. Torres-Labandeira and J. L. Vila-Jato, Eds.), p. 707. Kluwer Academic, Dordrecht, 1999. 24. N. J. Easton and S. F. Lincoln, “Modified Cyclodextrins, Scaffolds and Templates for Supramolecular Chemistry,” p. 293. Imperial College Press/World Scientific, London/Singapore, 1999. 25. T. Loftsson, in “Proc. Eleventh Int. Symp. on Cyclodextrins,” Reykjavik, 2002. 26. Cyclodextrin News (a monthly abstracting newsletter since 1986, dedicated exclusively to cyclodextrin literature, production, marketing, related conferences, etc.), edited and published by Cyclolab Ltd., Budapest.

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Deformation Behavior of Nanocrystalline Materials Alla V. Sergueeva, Amiya K. Mukherjee University of California, Davis, California, USA

CONTENTS 1. Introduction 2. Deformation Behavior 3. Main Issues 4. Concluding Remarks Glossary References

1. INTRODUCTION 1.1. Nanocrystalline Materials Nanostructure science and technology is a broad and interdisciplinary area of research, and development in this field has been growing very significantly worldwide in the past few years. It has the potential for revolutionizing the ways in which materials and products are created and the range and nature of functionalities that can be accessed. It is already having a significant commercial impact, which will assuredly increase in the future. Nanotechnology implies direct control of materials and devices on molecular and atomic scales, including fabrication of functional nanostructures with engineered properties, that takes advantage of physical, chemical, and biological principles that originate in the nanometer scale. New observed properties are attributed to a reduction in length scales to the point where known models for physical phenomena and interactions become larger than the actual size of the structure. New behavior at the nanoscale is not necessarily predictable from that observed at large size scales. The most important changes in behavior are caused not by the order of magnitude size reduction, but by newly observed phenomena intrinsic to or becoming predominant at the nanoscale, such as size confinement, predominance of interfacial phenomena, and quantum mechanics effects. Once it is possible to control feature size, it is also possible to enhance material properties and device functions beyond those that we currently know or even consider as feasible.

ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

Few reviews on the basic concepts, physics, structural features, properties, and applications of nanocrystalline materials (NCM) have been written in the past few years [1–8]. Many researchers in recent years have increasingly focused on the synthesis and processing of NCM, which have a great potential for functional applications (see next Section). Mechanical testing of NCM is characterized by features that are different from those documented for coarse-grained materials. The amount of the NCM produced by advanced processing methods is often small, and it is therefore difficult to prepare large bulk samples with nanocrystalline structures in comparison with those for conventional materials. The gauge length of testing specimens, unlike those used in testing coarse-grained materials, is usually on the order of several tens of micrometers to a few millimeters. However, due to the fact that the grain size is in scale of nanometers, it is anticipated that the experimental data obtained from these miniature specimens would still represent the bulk behavior of materials.

1.2. Severe Plastic Deformation There are many different methods developed for the production of NCM such as mechanical alloying [9], evaporation and condensation [6], precursor decomposition [10], powder consolidation [6], electrodeposition [11 12], sputter deposition [13 14], thermal or plasma spraying [15], crystallization from the amorphous state [16], and severe plastic deformation [17 18]. Reviews on some of these methods were made by Morris [8] and by Froes et al. [19]. Nevertheless, as it was determined by recent investigations [20 21], properties in NCM are very sensitive to their initial microstructures. Two grades of a nanocrystalline material (same composition and structure) that are prepared by different procedures and that have a similar grain size may exhibit different mechanical properties. On this basis, discrepancies may exist when comparing experimental data reported by investigators who used different processing techniques in producing the same nanocrystalline material.

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (305–316)

306 Moreover, it is very difficult to produce large bulk NCM that is free from porosity and other flaws and that has a wide range of grain size. Because of this difficulty, the dependence of strain rate on stress, grain size, and temperature in NCM is at present unclear and can be controversial. From the methods mentioned above, the technique of severe deformation under pressure (SePD) can produce bulk NCM free of porosity and surface flaws. This technique includes such methods as equal channel angular pressing (ECAP) [17], high pressure torsion (HPT) [17], and multiple forging (MF) [18 22], and has been intensively applied for producing of NCM. The current analysis on superplastic behavior is made for NCM produced mostly by HPT. The basic idea of torsion straining (Fig. 1) is to apply a very large pressure (5–7 GPa) to a sample and then twist to introduce large amount of torsion strain (up to true strain of 7) into the sample. The detailed description of this method can be found in [17].

1.3. Theoretical Predictions NCM materials contain a very large density of interfaces, causing them to differ structurally from crystals and glasses with the same chemical composition. Based on the classical scaling laws developed for conventional coarse-grained materials (grain size, d > 10 m), expected behavior would include very high yield strength, y , at low temperature due to constraints on dislocation motion (e.g., the Hall–Petch model predicts y ∝ d −1/2 ) and low strength/high ductility at higher temperatures due to short diffusion paths (e.g., the   strain rate, , in Coble creep scales as ∝ d −3 ). In addition, one might expect deviations from the classical scaling laws as grain size decreases. For example, when grain size becomes so very small, usual predictions based on dislocation theory may not be valid, and Hall–Petch behavior is expected to break down when the crystallite size becomes smaller than the average dislocation spacing or stable loop size. Diffusional flow might also deviate as the volume fraction of triple junctions increases. Past, experimental studies focused on (a) whether the Hall–Petch relation, which is well-established for coarsegrained materials, is valid for grain sizes in the nanocrystalline range [23]; (b) whether grain boundary diffusion creep [24], which is expected to be dominant at moderate temperatures (0.4–0.6Tm ), where Tm is the melting point

Figure 1. High pressure torsion (HPT).

Deformation Behavior of Nanocrystalline Materials

of the material, and low stresses, becomes significant at low temperatures (for example, room temperature); and (c) whether structural superplasticity [25], which has been observed at high temperatures (T > 05Tm ) and moderate strain rates (10−5 –10−2 s−1 , can be observed at low temperatures and high strain rates. There have been several reports on the variation of hardness with grain size that deviates from the normal Hall– Petch relationship [20 21]. Attempts have been made to calculate an approximate value for the grain size at which the Hall–Petch relationship breaks down (see for example [26]). Several derivations of the Hall–Petch equation are based on the concept of dislocation pile-ups at grain boundaries. (For a review of Hall–Petch theories, see for example [27].) However, transmission electron microscopy (TEM) evidence indicates that yielding involves the emission of dislocations from grain boundaries, rather than pileups against them [28 29]. As it was shown by Gryaznov and Trusov [30], parametric dependencies for the Hall–Petch relationship in NCM will be different from that for typical polycrystals, because the basic defects responsible for nanocrystal deformation are interface dislocations having Burgers vectors smaller than lattice dislocations. The large volume fraction of intercrystalline regions must strongly affect mechanical behavior. The nature of the structure of grain boundaries in NCM has been a controversial area. Early studies [31] proposed that interfaces in NCM represent a new class of solid-state structure lacking both shortand long-range orders. More recent studies [28 32 33] have disputed this view, and it is now concluded that the boundaries in NCM do not differ fundamentally in nature from those in coarser-grained materials. However, specific interface structures with a large amount of atomic disorder have been observed by Ranganathan et al. [33] in nanocrystalline palladium and titanium thin films. Theoretical discussion of nanomaterial at ambient temperature usually relates to the following: Starting with the concept of intragranular dislocation plasticity, as the grain size is progressively decreased, the formation of continuum dislocation pile-ups should become more difficult, and transition to discrete pile-up behavior will occur (end of Hall–Petch correlation). At smaller values of grain size, a pile-up approaches the limit of one dislocation. Then only an Orowan bowing type of mechanism is admissible. Finally, at some critical and yet lower length scale, no dislocations can be formed. Beyond this point, these authors believe [34] along with Masumura et al. [35] that plasticity, if it still exists, has to be interface driven. All of these transitions from traditional dislocation behavior can lead to a new grain size dependence of strength and plasticity, which may cause a fundamental change in deformation behavior of NCM. Intensive investigation of the deformation processes in NCM focuses on the main question of whether lattice dislocations exist and play the same role in nanograin interiors as with conventional coarse grains. As pointed out in the literature [36–38], the existence of dislocations in either free nanoparticles or nanograins composing a nanocrystalline aggregate is energetically unfavorable if their characteristic size, nanoparticle diameter, or grain size is lower than some critical size. And the critical size depends on such material characteristics as the shear modulus and the resistance to

307

Deformation Behavior of Nanocrystalline Materials

dislocation motion. The dislocation instability in nanovolumes is related to the effect of the so-called image forces occurring due to the elastic interaction between dislocations and either free surface of nanoparticle or grain boundaries adjacent to a nanograin [36–38]. The lack of mobile dislocations in nanograins has been well documented in electron microscopy experiments [21 39]. Moreover, it was suggested that mechanisms of plastic flow may be different in nanocrystals in comparison to usual polycrystals [40 41]. However, if the available literature concerning deformation behavior of NCM at room temperature was intensively investigated, we would find that very little is known about creep behavior and the actual diffusional accommodation at elevated temperatures [41]. At very low stresses and small grain sizes, early theoretical considerations indicated that vacancies, rather than dislocations, may be responsible for the production of creep strain. Two models, Nabarro– Herring creep [42 43] and Coble creep [24], have been formulated to account for such plastic flow. Nabarro–Herring creep [42 43] involves the diffusion of vacancies through the grain volume; the creep rate is given by   D Gb b 2   NH = ANH l (1) KT d G

1 nm. Based on this information, the triple junction process is likely to be important when the grain size is less than 10 nm [45]. Another important prediction for NCMs is their high ability for superplastic deformation. The ability of some materials to exhibit extensive plastic deformation, often without the formation of a neck prior to fracture, is generally known as structural superplasticity. Superplastic behavior is indicated in tension tests by large elongations, usually greater than 200 percent and sometimes in excess of 2000 percent. The two basic requirements for the observation of structural superplasticity are (a) a temperature greater than about onehalf of the melting temperature, Tm , and (b) a fine and equiaxed grain size (1 s−1 , which is comparable to conventional hot forming rates. In comparison, the superplastic strain rates of a material with a typical grain size of 15 m are 10−4 –10−3 s−1 [50]. The relationship between grain size and optimum strain rate of aluminum alloys is shown in Figure 2a (data taken from [50]) while Figure 2b shows the variation of superplastic temperature with grain size (data taken from [51 52]). From these figures, it is clear that by manipulating the grain size it is possible to (a) increase the superplastic strain rate and (b) decrease the superplastic temperature. The purpose of this chapter is to review the experimental data reported for deformation behavior of SePD processed NCM at elevated temperatures including our own results.

where ACo is a dimensionless constant and Dgb is the grain boundary diffusion coefficient. The Nabarro–Herring and Coble processes represent two independent and parallel mechanisms, so the faster process controls the creep behavior. Coble creep should predominate over Nabarro-Herring creep when   A b Dgb > Co (3) d Dl ANH This condition prevails at low homologous temperatures and very small normalized grain sizes. On this basis, it is expected that due to the ultrafine grain size and the large fraction of intergrain areas in NCM, Coble creep (diffusional creep along grain boundaries) would become important in these materials when crept at low temperatures. Room temperature diffusional creep, however, predicted to be appreciable in nanocrystalline samples, was not confirmed in subsequent investigation [44]. At the smallest grain sizes the significant volume associated with triple points must also become important [45]. However, according to a model proposed by Suryanarayana et al. [46], when the grain size is about 30 nm, which is very close to the grain size in Ni-P and Fe-B-Si used by Wang et al. [47], Deng et al. [48], and Xiao and Kong [49], the volume fraction of triple junctions is less than 1%, while the fraction of grain boundaries is about 10% by assuming grain boundary thickness of



2. DEFORMATION BEHAVIOR Because of specific microstructural features (very small grain sizes, high volume fraction of grain boundaries and triple points, lattice distortions, etc.), it is not surprising

308

Deformation Behavior of Nanocrystalline Materials 10 10

Strain Rate, s-1

10 10 10 10 10 10 10 10 10 10 10

7

6

5

T = 673 - 823 K

4

Nanocrystalline materials

3

2

1

0

-1

-2

-3

-4

a

-5

10

10

10

Mean Grain Size,µ m

Aluminum Alloys

10

10

10

10

-3

10

-2

10

-1

10

0

Mean grain size,µ m

10

1

10

2

3

- Al-4%Cu-0.5%Zr - Mg-1.5%Mn-0 .3%Ge - TiAl alloys

2

1

0

-1

-2

Nanocrystalline materials

b

-3

10 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

T/Tm Figure 2. The relationship between grain size and (a) optimum strain rate or (b) superplastic temperature demonstrates that by reducing the grain size it is possible to increase the strain rate or decrease the superplastic temperature (data taken from [50–52]). Both effects have technological significance.

that NCMs reveal specific variations of several fundamental physical parameters as compared to their large-grained counterparts, such as Curie and Debye temperatures, elastic modulus, diffusion coefficients [53], etc. Consequently, novel

physical and mechanical properties are obtained; for example, brittle ceramics may become ductile in nanocrystalline state [54]. However, it should be noted that data on mechanical properties of nanocrystals show contradictory trends. The possible reason for this is an expected difference in structure (porosity, density gradients, contaminations, internal stresses, defects including microcracks, etc.) of samples prepared in different laboratories even for the same grain size. Often this difference in structure is due to uncertain processing history and/or lack of proper control of process variables. Thus, most of the novel properties are found to be very sensitive to the procedure for sample preparation [20 21]. Thus, NCM produced by the powder sintering procedure did not demonstrate any superplasticity [55]. Moreover, among the NCM produced by other methods, some of them have shown the ability for superplastic flow, while others did not. The summary of our data and some additional literature data on deformation behavior of SePD NCM are given in Table 1. On the basis of the experimental results on SePD NCM, Mishra et al. [56] proposed an “exhaustion plasticity” mechanism. It is known that severe plastic deformation introduces a large population of dislocations into the specimen. In this “exhaustion plasticity” mechanism, the flow stress at the initial stage of deformation is not high enough to nucleate new dislocations for slip accommodation because of the very small value of grain size. The applied stress moves the preexisting dislocations, and grain boundary sliding is accommodated by the movement of these dislocations. As the easy paths of grain boundary sliding and dislocation movement get exhausted, the flow stress increases, which is indicated by a strain hardening stage observed in the superplastic deformation of the NCM. According to Mishra et al. [56], the peak flow stress can be viewed as the critical stress required to generate new dislocations during grain boundary sliding. Hence, the experimental results on commercially pure Ti and NiTi alloy have confirmed that nanostructure as well as excess of dislocations are not sufficient conditions for SePD materials to show superplasticity.

Table 1. A summary of observation of superplasticity in nanocrystalline materials.

Temp. ( C)

Stress at = 01 (MPa)

Max. stress (MPa)

Elong. (%)

70

7 × 10

−1

600

18



500

60

5 × 10−4

600

150



600

80

5 × 10−4

120

18



230

50

1 × 10−3

650 725

400 270

1530 790

380 560

Extensive strain hardening

Al-Mg-Li-Zr alloy (Mishra et al., 1998)

100

1 × 10−1

250 300

106 33

154 146

330 850

Extensive strain hardening

Ti-6 Al-4 V (Unpublished, 1998)

70

1 × 10−3

575

165



215

Material and reference b

Al-Ni-Mm (Higashi et al., 1993) Ti-6 Al-3.2 Mo (Salishchev et al., 1993) Zn-Al (Mishra et al., 1997) Ni3 Al alloy (Mishra et al., 1998)

a b

Gain sizea (nm)

Strain rate (s−1 )

Denotes the starting average grain size. This alloy showed significant coarsening during heating, so the microstructure during testing was not nanocrystalline.

Remarks

Stress at  = 40%

309

Deformation Behavior of Nanocrystalline Materials

2.1. Significant Differences 2.1.1. Excessive Strain Hardening Figure 3 shows flow curves of SePD 1420-Al and Ni3 Al alloys, commercially pure Ti, Ti-6Al-4V, and NiTi alloys. Excessive strain hardening was observed for all these materials when their grain size does not exceed 100 nm. Intensive grain growth at high temperatures (Fig. 3c and e) or low testing strain rate (Fig. 3d) leads to dramatic change in the deformation flow, and no strain hardening was observed. It is important to note that the strain hardening in the 1420-Al

a

True stress, MPa

2000

SePD 1420-Al alloy -1

-1

-1

-1

-2

-1

1 - 5x10 s

1

200

SePD Ni3Al

b

o

T = 300 C

250

True stress, MPa

300

-3

-1

Strain rate - 10 s

1500

2 - 1x10 s 3 - 1x10 s

1

o

1 - 650 C

1000

150

2

100

3

o

2 - 725 C 500

2

50 0 0.0

0

0.5

1.0

1.5

2.0

True strain

2.5

3.0

0

100

200

300

400

500

600

True strain 300

c

d

SePD CP Ti -4

Strain rate - 10 s

-1 o

1 - 450 C 300

1

o

2 - 550 C

200

2 100

0 0.0

0.2

0.4

0.6

True strain 1000

e

SePD oTi-6Al-4V alloy

T = 650 C

250

True stress, MPa

400

True stress, MPa

0.8

SePD TiNi

-3 -1

2 - 1x10 s

-4 -1

3 - 1x10 s

150

2

100 50

3

0 0.0

1.0

-2 -1

1 - 1x10 s

1

200

0.5

1.0

1.5

True strain

o

-5 -1

o

-2 -1

2.0

1 - 400 C; 1x10 s

2 - 650 C; 1x10 s

True stress, MPa

According to literature [57], the deformation of microcrystalline commercially pure Ti is associated with dislocation climb controlled by lattice diffusion. The stress exponent obtained by different investigators is close to 5, and activation energy is found to be 210–250 kJ/mol. This mechanism of deformation, namely five-power-law creep seems to be also operative in Ti in the nanocrystalline range at temperatures higher than 0.4Tm and at “conventional” strain rates (10−4 · · · 10−2 s−1 [12]. Apparent activation energy for both microcrystalline and nanocrystalline states seems to be higher than that for lattice diffusion reported for commercially pure Ti (153 kJ/mol [58]). In a NiTi alloy with nanostructure obtained by the same method (HPT), the deformation mechanism seems to be related to viscous glide of dislocations with a stress exponent of 3 [12]. Such a deformation mechanism is common for ordered structures. There is a lack of data on diffusion related processes in NiTi in the literature, but an approximate estimation of activation energy for self-diffusion gives a value of 220–230 kJ/mol [12]. So, the apparent activation energy of 210 kJ/mol calculated from the experimental data is in good agreement with such estimation, assuming that the activation energy for viscous glide, which is controlled by Darken interdiffusivity, is close to volume self-diffusivity in this case. Despite the fact that the grain size of this material is on the nanometer scale, and the volume fraction of grain boundaries is therefore significant, it seems that grain boundary sliding or operation of Coble creep are suppressed in NiTi. The results on NCM that did undergo superplastic deformation have established a number of general trends of the mechanical behavior of NCM at elevated temperatures such as excessive strain hardening during tensile tests, high flow stresses (compared to microcrystalline materials at comparable temperature and strain rate), a correlation between microstructural instability (in terms of onset of grain growth), and superplasticity, which will be discussed below. In microcrystalline superplasticity, the stress concentration that arises at grain boundary triple points (or ledges) due to grain boundary sliding is accommodated by diffusional processes or by (dislocation) slip processes. One of the fundamental issues in nanocrystalline superplasticity is: Which of the two accommodation processes is most important? In recent years, the possibility of superplasticity in NCM has been discussed in terms of diffusional flow. However, a grain boundary sliding mechanism with diffusional accommodation will not be able to explain all of the trends described above. These trends indicate that the mechanism of superplasticity in NCM is significantly different from the well-established mechanisms in microcrystalline materials.

800

1 600 400

2 200 0 0.0

0.2

0.4

0.6

True strain

0.8

Figure 3. Flow curves for (a) 1420-Al alloy at different strain rates, (b) Ni3 Al alloy at different temperatures, (c) CP Ti at different temperatures, (d) Ti-6Al-4V alloy at different strain rates, and (e) NiTi intermetallic at different temperatures show excessive strain hardening. Note the increase in strain hardening with increasing strain rate in 1420-Al alloy and Ti-6Al-4V alloy.

alloy (Fig. 3a) and Ti-6Al-4V alloy (Fig. 3d) increases with increasing strain rate. We have observed a similar trend in other SePD processed materials as well, for example, Zn-Al and 2124-Al alloys [59 60]. Similar elongations were reached in CP Ti (with initial grain size of about 120 nm) at 450 and 550  C and at a strain rate of 1 × 10−4 s−1 (157% and 166% respectively, but the shape of the flow curve has changed dramatically with decrease of test temperature (Fig. 3c). The flow curve at 450  C shows an extensive hardening period and stress dependence on strain (Fig. 3c, curve 1), whereas the curve at 550  C, when grain size was already more than 200 nm, has a shape typical for nearly steady state flow with relatively low stress flow (Fig. 3c, curve 2). These observations establish a trend. There is some grain growth in these NCMs during superplastic deformation. Because of the grain size dependence in the constitutive relationship for superplastic flow, strain hardening during superplasticity has been conventionally explained in terms of grain growth. However, such an explanation would not predict higher strain hardening at higher strain rate. A detailed microstructural investigation is required to explain this trend.

310

Deformation Behavior of Nanocrystalline Materials 1200

SePD Ni3Al-Cr alloy True stress, MPa

1000

o

-3

-1

1 - 50 nm; 725 oC; 1x10 -4s -1 2 - 6 µ m; 1050 C; 6x10 s

800 600

1

400 200

2 0 0.0

0.5

1.0

1.5

2.0

True strain Figure 4. A comparison of flow curves for Ni3 Al alloy in microcrystalline and nanocrystalline state. Note the higher flow stress in the nanocrystalline state, although both specimens showed similar ductility.

2.1.2. High Flow Stresses

σ /G

Figure 4 shows a comparison of the flow curves during superplasticity in a Ni3 Al alloy. This alloy was tested in both microcrystalline and nanocrystalline states. The high flow stresses during superplasticity of the nanocrystalline Ni3 Al alloy are apparent. Although the test temperature and grain size is different for the two microstructural conditions, the comparison is interesting because the overall ductility in both cases is similar. One of the features of conventional superplasticity in microcrystalline material is low flow stress. The observation of superplasticity with high flow stresses in NCM needs a new approach to explain the origin of such high flow stresses. First, the starting flow stress is around 170 MPa. Second, after significant strain hardening, the flow stress reaches the level of 780 MPa at 725  C. At 650  C, the flow stress values are even higher (Fig. 3b). Figure 5 shows the variation of normalized stress with temperature and grain size compensated strain rate for Ti-6Al-3.2Mo alloy. The raw stress-strain rate data were taken from Salishchev et al. [18 22]. Also, the data for a microcrystalline Ti-6Al-4V alloy [61] and the expected behavior from the grain boundary diffusion controlled mechanisms of deformation [62] are included. Titanium alloys 10

-1

10

-2

10

-3

10

-4

10

-5

10

-6

10

-7

10

-8

10

-9

10

-1 0

10

Ti-6Al-4X alloys

Nanocrystalline Ti-6Al-4Mo , 60 nm Sub-microcrystalline Ti-6Al-4Mo, 400 nm SePD Ti-6Al-4V, 120 nm Microcrystalline Ti-6Al-4V, 5µ m Grain Boundary Sliding -3

10

-2

10

-1

10

0

10 .

1

10

2

10

3

10

4

10

5

2

(εd )/DL Figure 5. The variation of normalized flow stress with temperature and grain size compensated strain rate for Ti alloys (data taken from [18 22 61 62]).

exhibit slip accommodated lattice-diffusion controlled grainboundary sliding. Two things are apparent from Figure 5: (a) the data for microcrystalline alloy agree well with the expected trend for slip accommodated lattice-diffusion controlled grain-boundary sliding, and (b) the data for the submicrocrystalline alloy and nanocrystalline material show higher flow stress. Higher flow stresses have been also observed for NCM, produced by other methods [63]. This suggests a transition in micromechanisms or additional difficulty in grain boundary sliding, which is opposite to some of the earlier expectations of enhanced superplasticity in NCM. One way to explain the origin of higher flow stresses for superplasticity in NCM is to consider the possible influence of grain size on slip accommodation during grain boundary sliding. Recently, Mishra and Mukherjee [64] have suggested that dislocation nucleation is difficult in NCM. A preliminary estimate of the stress required for nucleating such a dislocation can be obtained by using a relationship proposed by Hirth and Lothe [65] for dislocation generation by a Frank– Read source in the crystal lattice, which after simplification can be written as   b L = (4) ln − 167 4L1 −  b where L is the distance between the pinning points,  is the Poisson’s ratio, and  is the shear stress required to generate the dislocations. For superplastic deformation, the tensile stress needed to nucleate dislocation for slip accommodation can be √ calculated by approximating L = d/3,  = 033, and  = 3 . The choice of L = d/3 is based on the fact that the edge of the tetrakaidecahedron grain is one-third of the distance between the faces of the grains [66]. Equation (4) does not have (a) strain rate dependence (which is an integral part of high temperature deformation), (b) temperature dependence other than the modulus, and (c) the details of dislocation generation from grain boundaries. These effects can be significant, as the flow stress at high temperatures is known to be strain rate and temperature dependent in the superplastic region.

2.1.3. Parametric Dependencies In general, in order to understand the nature and origin of deformation mechanisms at high temperatures, information related to the dependence of strain rate on stress, temperature, and grain size is essential, and constitutive equations are very helpful in this case. For superplasticity, it is customary to evaluate the stress exponent, activation energy, and grain size exponent. Based on the values of these parametric dependencies, one can decide on the dominant micro mechanism of deformation. So far, there is very limited data on stress exponent and activation energy for superplastic flow in NCM. These data are listed in Table 2. We note that during some of the tests used to obtain the values mentioned below, significant grain growth occurred. The extent to which the results summarized below are influenced by concurrent grain growth is not clear. The stress exponent values for the alloys are close to the theoretically expected value of 2 for grain boundary sliding mechanisms. The activation energy in some cases is significantly higher than the activation energy for grain boundary diffusion. This is intriguing, because with

311

Deformation Behavior of Nanocrystalline Materials Table 2. Parametric dependencies for superplastic flow in nanocrystalline materials. Activation energya (kJ/mol)

Superplastic parameters Material and reference

Activation energy (kJ/mol)

For lattice diffusion

For grain boundary diffusion

151

142

84

2.3

315

150

97

∼25

90

142

84

5.8

270

150

97

∼25

ND

150

97

3

210

14b

44.9c

Stress exponent

Al-Ni-Mm alloy (Higashi et al., 1993) Ti-6 Al-3.2 Mo alloy (Salishchev et al., 1993) Al-Mg-Li-Zr alloy (Mishra et al., 1998) CP Ti (Sergueeva et al., 2001) Ti-6 Al-4V alloy (Sergueeva et al., 2001) NiTi (Sergueeva et al., 2001)

∼2

a The activation energy values are for self diffusion of matrix element and these values are taken from Deformation Mechanism Maps by Frost and Ashby (1982). b The activation energy for diffusion of Ni in Ti (taken from Smithells Metals Reference book, 7th ed., 1992). c The activation energy for grain boundary diffusion in Ni (taken from Smithells Metals Reference Book, 7th ed., 1992).

the abundance of grain boundaries in NCM, an activation energy for grain boundary related deformation mechanism is expected at elevated temperatures.

2.1.4. Effect of Short Annealing A unique experimental observation is related to the effect of a short anneal on the flow behavior. An example of this is shown in Figure 6. The annealing of SePD processed nanocrystalline Ni3 Al does not change the grain size significantly, but it has a dramatic effect on the flow behavior. The initial flow stress increases and the shape of the flow curve changes. The overall ductility reduces from 350% in the as received condition to ∼100% after the short anneal. It should be noted, that a strong effect of the short annealing of SePD Ti was revealed at room temperature tensile tests. Thus, short annealing of SePD Ti at 300  C results in a 30% increase in strength combined with adequate plasticity (Fig. 7, curve 2) as compared to as-deformed state

(Fig. 7, curve 1) [67]. Annealing for a longer time even at lower temperature decreases strength (Fig. 7, curve 3). A number of tensile tests of severe plastic deformed CP Ti after annealing at different temperatures for 10 min revealed a strong dependence of strength characteristics on annealing temperature with a peak in the range of 250  C to 300  C. Microstructural analysis revealed a formation of specific ordered structure in grain boundary regions (Fig. 8) [68], which was not observed in as-deformed SePD Ti. In the sample, which has demonstrated the highest strength, about 80% of grain boundaries have a such special structure.

2.1.5. Correlation Between Microstructural Instability and Superplasticity The driving force for grain growth is quite high in NCM because of the high interfacial area per unit volume. We have observed a nice correlation between the temperature for microstructural instability and onset of superplasticity in

2500

SePD Ni3 Al

1400

-3 -1

2000

1

1500

1 - as-processed 2 - annealed o at 750 C; 1 min

2 1000

500

1200

-3

-1

Tested at 20 C, 10 s

2 3

1000

1 800 600 400

1 - as-processed o

2 - annealed at 300 C for 10 min

200

0

SePD CP Ti o

True stress, MPa

True stress, MPa

o

650 C; 1x10 s

0

100

200

300

400

Elongation, % Figure 6. Effect of a short anneal on flow behavior of a nanocrystalline Ni3 Al alloy.

0 0.0

o

3 - annealed at 250 C for 1h

0.1

0.2

0.3

True strain Figure 7. True stress-true strain curves of CP Ti after HPT and annealing.

312

Deformation Behavior of Nanocrystalline Materials

a

1420-Al .

1200

DSC curve -3 -1 Elongations at ε =1x10 s

0.1

900

0.0

600

Elongation, %

dH/dt, W/g exothermic endothermic

0.2

300

-0.1 onset of primary grain growth

-0.2

0

100

200

300

400

o

500

0

Temperature, C

NCM. In some ways this is to be expected because both grain boundary migration and grain boundary sliding involve diffusion. An important implication of this is that one needs to establish the narrow temperature range in which there is enough thermal activation for grain boundary sliding but the grain growth rate is relatively low. This is important for observing the grain boundary sliding regime without significant influence of grain growth, particularly from the viewpoint of obtaining parametric dependencies for constitutive modeling. We are investigating the microstructural instability by three different methods: differential scanning calorimetry (DSC), in-situ TEM heating, and TEM of heat treated specimens. The correlation between the DSC signal for thermal stability and increase in ductility is shown for SePD processed 1420-Al and NiTi alloys in Figure 9. The onset of grain growth was independently established by analyzing the video recording of in-situ TEM heating. The results show that the DSC can work as a very good tool to establish the optimum superplastic temperature for NCM.

2.1.6. Microstructural Observations We have conducted two types of TEM investigations: (a) microstructural investigation of thin foils at room temperature and (b) video recording of microstructural changes during in-situ heating. We have established that the present results at some temperatures reflect superplasticity in the truly nanocrystalline state, as the microstructure remained nanocrystalline during the entire test. For example, Figure 10 shows the microstructure of the nanocrystalline Ni3 Al alloy after testing at 650  C to 380% elongation. In general the dislocation activity appears to quite limited. A few lattice dislocations were visible, as marked in Figure 10. It is interesting to note that dislocation 2 has one end in the grain boundary. The appearance of these types of dislocations suggests the grain boundaries as sources of dislocations. This would be consistent with slip accommodation during grain boundary sliding. The 1420-Al alloy showed high strain rate superplasticity at a relatively low temperature where the nanostucture was relatively stable. It is necessary to note that in the earlier studies [69 70]

300

40

b DSC curve -3 -1 Elongations at ε =1x10 s . 250

30

200

20

150

10

100

NiTi

Elongation, %

Figure 8. HREM of GB region in CP Ti after HPT and annealing at 250  C for 10 min.

dH/dt, W/g exothermic endothermic

50

50

0

onset of precipitation and grain coarsening

-10

25 onset of primary grain growth

-20

0

100

200

300

400

500

600

o

700

800

Temperature, C

Figure 9. Combined plots showing variation of DSC signal and elongation with temperature for SePD (a) 1420-Al alloy and (b) NiTi.

intensive grain growth prior to testing or during testing to the submicrocrystalline range was detected. Microstructural investigation was also conducted of nanocrystalline commercially pure Ti, which showed high elongation but with parametric dependencies different from that for classical superplasticity (Table 2). This material has demonstrated a five-power-law creep that associated with dislocation climb controlled by lattice diffusion. The enhanced plasticity observed in this pure metal and TEM

Figure 10. A bright field transmission electron micrograph of a Ni3 Al specimen superplastically deformed at 650  C to an elongation of 380%. Some of the lattice dislocations are marked by arrows.

313

Deformation Behavior of Nanocrystalline Materials

observations of the structure after deformation suggest that grain boundary sliding also took place during deformation at elevated temperatures. Figure 11 presents the microstructure of Ti produced by severe plastic deformation and deformed at 350  C (Fig. 11a and b) and 600  C (Fig. 11c, d, and e) to strain of 65% and 190%, respectively. It is seen that despite such high elongations the grains remain equaxied in both cases. Twinning was not observed during deformation at 350  C (Fig. 11a and b), while at 600  C twins were formed (Fig. 11c). Moreover, some twins emerging from grain boundaries were revealed (Fig. 11d). Such twinning can be formed as a result of grain boundary sliding in order to resolve stress concentration on grain boundary. So, during elevated temperature deformation there seems to be a competition between these two mechanisms and both of them play a role in the deformation process. As shown by analysis of stress-strain curves, the rate controlling deformation mechanism in investigated range of strain rates and temperatures is dislocation glide controlled by dislocation climb. High dislocation activity should be noted in the deformed structures (Fig. 11a and d). In the sample deformed at higher temperature, stacking faults have been also observed (Fig. 11e). Some of them look like regular extended dislocation (marked as “A”), whereas the others

have triangular shape and might be formed by dissociation of stair rod dislocations (marked as “B”).

2.1.7. Absence of Cavitations Microcrystalline superplastic materials have shown clear evidence for intergranular cavitation. This includes observation on Al-Cu-Li-Zr alloy [71] and Ni3 Al alloy [72]. However, current investigation on NCM fails to reveal any evidence of cavitation. A critical question may be: Can one maintain the stress concentration at grain boundary triple points, grain boundary ledges, or at particles on the grain boundary long enough so that cavity nucleation can take place? Or can it be that diffusional relaxation in NCM is so fast that one cannot maintain that condition? In an analysis of cavity nucleation in superplasticity [73], the authors have proposed a competition between the characteristic relaxation time for grain boundary sliding leading to the development of stress concentration at ledges. Cavities may nucleate at these sites, if the incubation period for cavity nucleation is less than the characteristic time for the relaxation of stress concentration by localized diffusion creep. The analysis yields a quantitative cavity nucleation map in terms of normalized stress versus normalized cavity height/spacing. The analysis also suggests that in such ultrafine grain materials, it is possible for cavities to nucleate under some rather limited experimental conditions.

3. MAIN ISSUES 3.1. Effect of Processing

Figure 11. Microstructure of CP Ti produced by severe plastic deformation and deformed at 350 [(a) bright field; (b) dark field images] and at 600  C [(c)–(e): (d) is enlarged image of the selected area on (c); (e) is enlarged image of the selected area on (d).]

On the basis of the experimental results on SePD NCM, Mishra et al. [56] proposed an “exhaustion plasticity” mechanism. It is known that severe plastic deformation introduces a large population of dislocations into the specimen. In this “exhaustion plasticity” mechanism, the flow stress at the initial stage of deformation is not high enough to nucleate new dislocations for slip accommodation because of the very small value of grain size. The applied stress moves the preexisting dislocations, and grain boundary sliding is accommodated by the movement of these dislocations. As the easy paths of grain boundary sliding and dislocation movement get exhausted, the flow stress increases, which is indicated by a strain hardening stage observed in the superplastic deformation of the NCM. According to Mishra et al. [56], the peak flow stress can be viewed as the critical stress required to generate new dislocations during grain boundary sliding. Mishra et al. [56] further predicted that the NCM prepared by sintering of nanocrystalline powders would not exhibit large tensile plasticity due to the lack of the large population of pre-existing dislocations. From the evaluation of the behavior of NCM, it is clear that a nanocrystalline grain size does not appear to be a sufficient condition for superplasticity. A number of pure metals and alloys processed by severe plastic deformation to grain sizes alcohols > hydrophobic dosants. This response order is more pronounced for the G4–G8-modified surfaces and is dictated by the PAMAM structure which possesses hydrogen-bonding exoreceptors and endoreceptors. The G4-modified surface is the most responsive material probably; although it is the smallest of the spheroidal dendrimers, its interior endoreceptors are most accessible. G0 and G2 dendrimer films are not as effective at discriminating between the three different classes of probes since these surfaces have few or no free amine terminal groups and no coherent endoreceptors ability. An atomic force microscopy investigation of PAMAM G4 and G8 dendrimers adsorbed on Au surfaces has been carried out [69]. Heights measured for isolated, adsorbed dendrimers indicate that they are substantially more oblate than expected from their spherical shapes in solution. By controlling dendrimer concentration and exposure time during adsorption, modified surfaces ranging from isolated molecules to near-monolayer coverage are obtained. Exposure of surfaces bearing adsorbed isolated dendrimers to hexadecanethiol solutions changes their conformation from oblate to prolate as more stable thiol–Au bonds replace some of the amine– Au bond. Surfaces of near-monolayer coverage gradually agglomerate, forming dendrimer “pillars” up to 30 nm high. Vapor-sensitive thin-film resistors comprising gold nanoparticles and different types of dendrimers such as polyphenylene, poly(propyleneimine), and poly(amidoamine) have been prepared via layer-by-layer self-assembly and characterized by UV-vis spectroscopy, atomic force microscopy, and conductivity measurements [70]. Dosing the films with vapors of toluene, 1-propanol, and water significantly increases the film resistances. It is demonstrated that the chemical selectivity of this response is controlled by the solubility properties of the dendrimers. Interactions between alkanethiols and dendrimer–gold nanocomposites in ethyl acetate have been investigated by adding alkanethiols such as 1-decanethiol and 1,10decanedithiol into PAMAM dendrimers with surface methyl ester group–gold nanocomposites [71]. In the case of PAMAM G1.5 dendrimer–gold nanocomposites (diameter: 2.1 nm), the addition of 1-decanethiol leads to smaller gold particles, while that of 1,10-decanedithiols provides the formation of large, highly regular, spherical clusters with diameters of 40–60 nm. On the other hand, in the case

Dendrimer–Metal Nanocomposites

of PAMAM G5.5 dendrimer–gold nanocomposites (diameter: 2.4 nm), the addition of 1-decanethiol shows a distinct tendency into smaller and larger spheres, whereas that of 1,10-decanedithiols gives irregular spherical clusters with diameters of 30 nm. Such morphology changes by addition of alkanethiols occur due to stronger interaction between gold and alkanethiols than that between PAMAM dendrimer and gold. Finally, bioapplications of dendrimer–metal nanocomposites are briefly described. Since PAMAM dendrimers are able to solubilize many organic materials or inorganic materials, bioactive materials may be combined in one nanoscopic delivery vehicle such as dendrimer. Studies using antibody/dendrimer conjugates in vitro and in vivo in experimental animals have demonstrated these conjugates to be nontoxic and able to target biologic agents to specific cells [72, 73]. Dendrimer–silver nanocomposites display antimicrobial activity comparable to or better than those of silver nitrate solutions [74]. Interestingly, increased antimicrobial activity is observed with dendrimer carboxylate salts. This effect may be caused by the very high local concentration of nanoscopic size silver composite particles that are accessible for microorganisms. PAMAM dendrimers have been used for imaging but only after appropriate surface modification. The potential of PAMAM–metal chelate–antibody constructs in radioimmunotherapy and imaging has been recognized [75]. Generally, fluorescent markers, such as fluorescein, or chelated metal ions, such as Gd3+ for magnetic resonance imaging, are used to generate a measurable signal [76, 77]. Radioactive dendrimer nanocomposites also have the potential to be delivered to the tumor.

GLOSSARY Dendrimer Macromolecules with a regular and highly branched three-dimensional architecture. Dendrimer–metal nanocomposite Organic–inorganic hybrid structure consisting of metal nanoparticles in the internal or external region of dendrimer. Plasmon resonance band of gold An extinction maximum band due to the dipole resonance of small gold spheres. Transmission electron microscopy Transmitted-light microscopy using an electron beam and magnetic lenses to produce a magnified image on a fluorescent screen.

REFERENCES 1. E. Buhleier, W. Wehner, and F. Vögtle, Synthesis 155 (1978). 2. R. G. Denkewalter, J. F. Kolc, and W. J. Lukasavage, U. S. Patent 4, 410, 688, 1979; 4, 289, 872, 1981; 4, 360, 646, 1982; 4, 410, 688, 1983. 3. D. A. Tomalia, H. Baker, J. Dewald, M. Hall, G. Kallos, S. Martin, J. Roeck, J. Ryder, and P. Smith, Polym. J. 17, 117 (1985). 4. G. R. Newkome, Z. Yao, G. R. Baker, and V. K. Guota, J. Org. Chem. 50, 2004 (1985). 5. C. J. Hawker and J. M. J. Fréchet, J. Am. Chem. Soc. 112, 7638 (1990). 6. J. M. J. Fréchet, Science 263, 1710 (1994). 7. G. R. Newkome, E. He, and C. N. Moorefield, Chem. Rev. 99, 1689 (1999).

325 8. “Dendrimers II: Architecture, Nanostructure and Supramolecular Chemistry” (F. Vögtle, Ed.), Topics in Current Chemistry, Vol. 210. Springer-Verlag, Berlin, 2000. 9. F. Zeng and S. C. Zimmermann, Chem. Rev. 97, 1681 (1997). 10. A. W. Bosman, H. M. Janssen, and E. W. Meijer, Chem. Rev. 99, 1665 (1999). 11. D. Astruc and F. Chardac, Chem. Rev. 101, 2991 (2001). 12. C. J. Jackson, H. D. Chanzy, F. P. Booy, B. J. Drake, D. A. Tomalia, B. J. Bauer, and E. J. Amis, Macromolecules 31, 6259 (1998). 13. R. M. Crooks, B. I. Lemon III, L. K. Yeung, and M. Zhao, in “Topics in Current Chemistry” (F. Vögtle, Ed.), pp. 81–135. Springer, Heidelberg, 2000. 14. “Clusters and Colloids” (G. Schmid, Ed.), VCH Verlag, Weinheim, 1994. 15. Y. Xia, J. A. Rogers, K. E. Paul, and G. M. Whitesides, Chem. Rev. 99, 1823 (1999). 16. A. P. Aliviasatos, Science 271, 933 (1996). 17. L. L. Beecoft and C. K. Ober, Chem. Mater. 9, 1302 (1997). 18. X. K. Zhao, S. Barai, R. Rolandi, and J. H. Fendler, J. Am. Chem. Soc. 110, 1012 (1988). 19. N. Toshima, M. Harada, T. Yonezawa, K. Kushihashi, and K. Asakura, J. Phys. Chem. 95, 7448 (1991). 20. K. Torigoe, Y. Nakajima, and K. Esumi, J. Phys. Chem. 97, 8304 (1993). 21. K. Esumi, A. Suzuki, N. Aihara, K. Usui, and K. Torigoe, Langmuir 14, 3157 (1998). 22. M. Zhao, L. Sun, and R. M. Crooks, J. Am. Chem. Soc. 120, 4877 (1998). 23. L. Bologh and D. A. Tomalia, J. Am. Chem. Soc. 120, 7355 (1998). 24. M. Zhao and R. M. Crooks, Angew. Chem. Int. Ed. 38, 364 (1999). 25. M. Zhao and R. M. Crooks, Adv. Mater. 11, 217 (1999). 26. M. Zhao and R. M. Crooks, Chem. Mater. 11, 3379 (1999). 27. M. E. Garcia, L. A. Baker, and R. M. Crooks, Anal. Chem. 71, 256 (1999). 28. V. Chechik, M. Zhao, and R. M. Crooks, J. Am. Chem. Soc. 121, 4910 (1999). 29. V. Chechik and R. M. Crooks, J. Am. Chem. Soc. 122, 1243 (2000). 30. K. Esumi, A. Suzuki, A. Yamahira, and K. Torigoe, Langmuir 16, 2604 (2000). 31. K. Esumi, R. Nakamura, A. Suzuki, and K. Torigoe, Langmuir 16, 7842 (2000). 32. F. Gröhn, B. J. Bauer, Y. A. Akpalu, C. L. Jackson, and E. J. Amis, Macromolecules 33, 6042 (2000). 33. L. Bologh, R. Valluzzi, K. S. Laverdure, S. P. Gido, G. L. Hagnauer, and D. A. Tomalia, J. Nanoparticles Res. 1, 353 (1999). 34. K. Esumi, T. Hosoya, A. Suzuki, and K. Torigoe, J. Colloid Interface Sci. 226, 346 (2000). 35. K. Esumi, T. Hosoya, A. Suzuki, and K. Torigoe, J. Colloid Interface Sci. 229, 303 (2000). 36. K. Torigoe, A. Suzuki, and K. Esumi, J. Colloid Interface Sci. 241, 346 (2001). 37. A. Manna, T. Imae, K. Aoi, M. Okada, and T. Yogo, Chem. Mater. 13, 1674 (2001). 38. S. Kéki, J. Török, G. Deák, L. Daróczi, and M. Zsuga, J. Colloid Interface Sci. 229, 550 (2000). 39. K. Esumi, A. Kameo, A. Suzuki, and K. Torigoe, Colloids Surf. A 189, 155 (2001). 40. R. M. Crooks, M. Zhao, L. Sun, V. Chechik, and L. K. Yeung, Acc. Chem. Res. 34, 181 (2001). 41. K. Esumi and K. Torigoe, Progr. Colloid Polym. Sci. 117, 80 (2001). 42. P. N. Floriano, C. O. Noble IV, J. M. Schoonmaker, E. D. Poliakoff, and R. L. McCarley, J. Am. Chem. Soc. 123, 10545 (2001). 43. F. Gröhn, G. Kim, B. J. Bauer, and E. J. Amis, Macromolecules 34, 2179 (2001).

326 44. J. Zheng, M. S. Stevenson, R. S. Hikida, and P. G. Van Patten, J. Phys. Chem. B 106, 1252 (2002). 45. A. Hierlemann, J. K. Campbell, L. A. Baker, R. M. Crooks, and A. J. Ricco, J. Am. Chem. Soc. 120, 5323 (1998). 46. V. V. Tsuruk, F. Rinderspacher, and V. N. Bliznyuk, Langmuir 13, 2171 (1997). 47. Y. Sayer-Sweet, D. M. Hedstrand, R. Spinder, and D. A. Tomalia, J. Mater. Chem. 7, 1199 (1997). 48. Y.-S. Seo, K.-S. Kim, K. Shin, H. White, M. Rafailovich, J. Sokolov, B. Lin, H. J. Kim, C. Zhang, and L. Balogh, Langmuir 18, 5927 (2002). 49. M.-K. Kim, Y.-M. Jeon, W. S. Jeon, H.-J. Kim, S. G. Hong, C. G. Park, and K. Kim, Chem. Commun. 667 (2001). 50. L. K. Yeung and R. M. Crooks, Nano Lett. 1, 14 (2001). 51. Y. Li and A. El-Sayed, J. Phys. Chem. B 105, 8938 (2001). 52. E. H. Rahim, F. S. Kamounah, J. Frederiksen, and J. B. Christensen, Nano Lett. 1, 449 (2001). 53. T. Mizugaki, M. Ooe, K. Ebitani, and K. Kaneda, J. Mol. Catal. A 145, 329 (1999). 54. M. Ooe, M. Murata, T. Mizugaki, K. Ebitani, and K. Kaneda, Nano Lett., 2, 999 (2002). 55. Y. Niu, L. K. Yeung, and R. M. Crooks, J. Am. Chem. Soc. 123, 6840 (2001). 56. I. T. Horvath and J. Rabai, Science 266, 72 (1984). 57. B. Cornils, Angew. Chem. Int. Eg. Engl. 36, 2057 (1997). 58. D. P. Curran, Angew. Chem. Int. Eg. Engl. 37, 1174 (1998). 59. L. K. Yeung, C. Lee, K. P. Johnston, and R. M. Crooks, Chem. Commun. 2290 (2001). 60. K. Esumi, K. Satoh, A. Suzuki, and K. Torigoe, J. Jpn. Soc. Colour Mater. 73, 434 (2000). 61. U. Kreibig and M. Vollmer, “Optical Properties of Metal Clusters.” Springer, Berlin, 1995.

Dendrimer–Metal Nanocomposites 62. C. F. Bohren and D. R. Huffman, “Absorption and Scattering of Light by Small Particles.” Wiley-Interscience, New York, 1983. 63. R. G. Ispasoiu, L. Balogh, O. P. Varnavski, D. A. Tomalia, and T. Goodson III, J. Am. Chem. Soc. 122, 11005 (2000). 64. O. Varnavski, R. G. Ispasoiu, L. Balogh, D. A. Tomalia, and T. Goodson III, J. Chem. Phys. 114, 1962 (2001). 65. J. Y. Ye, L. Balogh, and T. B. Norris, Appl. Phys. Lett. 80, 1713 (2002). 66. J.-A. He, R. Valluzzi, K. Yang, T. Dolukhanyan, C. Sung, J. Kumar, S. K. Tripathy, L. Samuelson, L. Balogh, and D. A. Tomalia, Chem. Mater. 11, 3268 (1999). 67. S. Watanabe and S. Regen, J. Am. Chem. Soc. 116, 8855 (1994). 68. H. Tokuhisa, M. Zhao, L. A. Baker, Vy T. Phan, D. L. Dermody, M. E. Garcia, R. F. Peez, R. M. Crooks, and T. M. Mayer, J. Am. Chem. Soc. 120, 4492 (1998). 69. A. Hierlemann, J. K. Campbell, L. A. Baker, R. M. Crooks, and A. J. Ricco, J. Am. Chem. Soc. 120, 5323 (1998). 70. N. Krasteva, I. Besnard, B. Guse, R. E. Bauer, K. Mullen, A. Yasuda, and T. Vossmeyer, Nano Lett. 2, 551 (2002). 71. K. Esumi, K. Satoh, and K. Torigoe, Langmuir 17, 6860 (2001). 72. D. Page and R. Roy, Bioconjugate Chem. 8, 714 (1997). 73. A. Bielinska, J. F. Kukowska-Latallo, J. Johnson, D. A. Tomalia, and J. R. Baker, Jr., Nucl. Acids Res. 24, 2176 (1996). 74. L. Balogh, D. R. Swanson, D. A. Tomalia, G. L. Hagnauer, and A. T. McManus, Nano Lett. 1, 18 (2001). 75. C. Wu, M. W. Brechbiel, R. W. Kozak, O. A. Gansow, Bioorganic Med. Chem. Lett. 4, 449 (1994). 76. L. D. Margerum, B. K. Campion, M. Koo, N. Shargill, J. Lai, A. Marumoto, and P. C. Sontum, J. Alloys Compd. 249, 185 (1997). 77. Y. Kim and S. C. Zimmerman, Curr. Opin. Chem. Biol. 2, 733 (1998).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Dendritic Encapsulation Christopher S. Cameron, Christopher B. Gorman North Carolina State University, Raleigh, North Carolina, USA

CONTENTS 1. Introduction 2. Encapsulation 3. Conclusions Glossary References

the establishment of a barrier to control the access of some system to its surroundings. Dendrimers represent a way to accomplish encapsulation on the molecular scale. Molecules with potentially useful physical and/or chemical properties can be encapsulated as the core of dendrimers. Ideally, dendritic encapsulation will result in new properties for the core molecules and ultimately new materials. This is the essence of materials research.

1. INTRODUCTION

1.1. Synthesis of Dendrimers

The search for new materials with unusual properties is a vital area of research. Since the original reported syntheses by Newkome et al. [1] and Tomalia et al. [2], dendrimers have assumed a prominent role in this search for new materials. In the past four years, this role has grown rapidly as evidenced by the large numbers of papers appearing in literature since 1998. A cursory review of cited journals indicates the importance and level of interest in the scientific community concerning dendrimer research. The research on dendrimers is now a multidiscipline effort involving synthetic chemistry, inorganic chemistry, organometallic chemistry, analytical chemistry, and polymer science. This level of scientific interest is driven by several properties of dendrimers. Among these are an unusual melt viscosity due to the hyperbranched structure (as described below), unusual properties when studied as thin films on surfaces, the ability to hold catalysts and influence the catalytic properties of catalysts within them, an emerging ability to foster cavities within them that can be used to bind small molecules, and the ability to mimic biomolecules and potential for drug and gene delivery. Several of these properties derive from the encapsulating abilities of dendrimers. The phenomenon of encapsulation can be observed in a multitude of different situations from atomic to macro scale. Medicines are regularly encapsulated within water-soluble capsules for delivery. Life exists because the components of cells are encapsulated by plasma membranes. Enzymes exhibit remarkable substrate binding specificity and perform remarkable organic chemistry in aqueous environments because active sites are encapsulated by the protein structure. Power cords are encapsulated by insulating material so that electric current can be channeled to power computers and lab equipment. In the broadest terms, encapsulation is

Dendrimers are a class of macromolecules consisting of one or more polymeric, regularly branching arms, known as dendrons, which radiate from a central core (Fig. 1). They are synthesized in a stepwise iterative fashion which follows either divergent or convergent methodology [3]. In the divergent methodology, first used by Newkome et al. and Tomalia et al. [1, 2], the core molecule serves as a polyvalent initiation point and the dendrons are grown outward from the core in a repetitive activate/couple cycle. Each time a coupling cycle occurs, a new layer of branch points, referred to as a “generation,” is added to the growing dendrimer. The divergent methodology is fast, efficient, and can yield large dendrimers. Its principal disadvantage is that the yield of complete coupling decreases as the generation and thus the number of reaction sites increases, creating some polydispersity in the molecular weight of the final product. In the convergent methodology, first reported by Hawker and Fréchet [4], the dendrons are synthesized by starting at the tips, which will be the periphery of the dendrimer, and working backwards to a single reactive functional group. Again, each time a coupling cycle occurs, a new layer of branch points is added and the “generation” of the dendrimer increases. Once the dendrons are fully synthesized, they are linked to the core in a final activate/couple cycle. Monodisperse dendrimers can be obtained this way, but the purification after each step renders a practical limit to the size of the molecules that can be synthesized. Furthermore, steric congestion may limit the efficiency of coupling in this approach as well. The essential difference between the two methodologies is in the fashion in which the dendrimer is assembled. In the divergent methodology, as soon as the activate/couple cycle begins, an actual dendrimer is synthesized and subsequent repetitions of the activate/couple cycle

ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (327–335)

328

Dendritic Encapsulation

1.2. Advantages of Dendrimers

Figure 1. Conceptual Dendrimer—The relative spatial orientation of the various parts of the dendrimer are shown. The central core is attached to the dendrons, which define the nanoenvironment around the core. The peripheral functional groups define the solubility of the dendrimer and/or provide multiple reactive sites for chemistry. The dendrimer shown here has three layers of branch points in the interior and would thus be designated a third-generation dendrimer.

As stated previously, encapsulation begins by surrounding a core molecule with a material. This effect could be achieved using conventional polymers. However, studies comparing dendrimer versus polymeric encapsulation have shown considerable differences between the two systems [14]. Structurally, dendrimers are composed of a core moiety surrounded by dendrons radiating from the core (Fig. 1). Such a structural arrangement predisposes dendrimers towards encapsulation when compared to linear polymers. There is substantial precedent for the synthesis of dendrimers as monodisperse or nearly monodisperse species. Linear polymers could be synthesized in the same, stepwise fashion, but protocols to isolate and to purify these types of molecules are less precedented. The mode of synthesis of dendrimers also facilitates the installation of different functionalities at the center and periphery. Dendrimers are modular in construction and composition, thus allowing the core, the interior dendrimer nanoenvironment, and the dendrimer peripheral functional groups to be changed independently of each other with relative ease [15]. The value of the “modular” nature of dendrimers and the potential applications for encapsulation lead to a number of possibilities for design of new functional molecules [16].

1.3. Dendrimer Cores serve to increase the size of the dendrimer. With the convergent methodology, a dendrimer is not actually assembled until dendrons of a desired size have been synthesized and then coupled to a core. Numerous reviews and articles have thoroughly covered dendrimer synthesis and the many types of dendrimers that have been prepared using both methodologies [3, 5–12]. Dendrimers are particularly well suited for the function of encapsulation (Fig. 1). When a dendrimer becomes sufficiently large, normally measured by generation number, it adopts a globular shape. This globular shape is the natural result of an architecture in which large bulky dendrons are focally attached to a central core. DeGennes dense packing, sterics, and simple mass effect cause the dendrons to envelope the core molecule, thus leading to encapsulation. While encapsulation by dendrimers can lead to alteration of physical and chemical properties, there is some scientific and semantic difficulty in defining the critical size or generation or molecular weight required to effect this change [13]. The term “site isolation” is often used when encapsulation is determined to be complete; this term is subject to interpretation. For dendrimers, it is possible to encapsulate many different types of core molecules in different types of “nanoenvironments” for the purpose of understanding structure–property relationships. A basic understanding of structure–property relationships is the key to the development of new materials with unusual properties and is one of the core tenets of scientific research. This ultimately is the basis for the broad-based, multidisciplinary interest in dendrimers.

Dendrimer cores can be classified as “nonfunctional” or “functional.” “Nonfunctional” cores are molecules that have no specifically interesting chemical or physical properties; and serve only as a polyvalent focal point of the dendrimer. Dendrimers assembled with “nonfunctional” cores either focus on materials applications other than encapsulation [17], such as supramolecular constructs [18] and selfassembly processes [19–23] or utilize the dendrimer as a polyvalent scaffold, as in DNA transfection [7, 24, 25], drug delivery [26–28], or even soluble solid-state synthesis [29]. In some of these applications, the dendrimer is used for encapsulation of small molecule guests. This type of non-covalent encapsulation is interesting and potentially useful, especially for extraction, solubility, and drug delivery. It is not, however, treated in any detail here. A second class of encapsulating dendrimers is based on functional cores. “Functional” cores are moieties which, in addition to serving as the polyvalent focal point for the dendrimer, have some observable and potentially useful physical and/or chemical property. For dendrons containing a “functional” core, encapsulation is the predominant research focus. Observable physical/chemical properties fall into three broad categories [30]: photophysics (absorbance, luminescence, fluorescence, and phosphorescence), electrochemistry (kinetic electron transfer rate and thermodynamic redox potential) [9, 31, 32], and ligand binding (ligands, gas binding, and catalysis) [6, 8, 33, 34]. A wide variety of molecules have been employed as “functional” cores (Table 1). Most can be classified in one of four categories: porphyrins/phthalocyanines, inorganic clusters, ionic complexes/organometallics, and fluorescent dyes. Porphyrin and phthalocyanine moieties [30, 35, 36] are especially useful because they exhibit three classes

329

Dendritic Encapsulation

of observable properties: photophysics, electrochemistry, and ligand binding. Many dendrimers have been synthesized with porphyrin/phthalocyanine cores. Several different types of inorganic clusters have served as cores. One impetus behind the use of clusters as cores is ligand binding and when mimicking and studying enzyme active sites [37, 38]. Clusters that have been used include iron-sulfur, Fe4 S4 [39], rhenium-selenide, Re6 Se8 [40, 41], molybdenum-chloride, Mo6 Cl8 [42], oxymolybdo-sulfur, MoOS4 [37], and metal nanoparticles [34, 43]. Ionic complexes and organometallics [44, 45] have been popular because of the well-established synthetic conditions by which a dendrimer may be assembled around such a core. Examples of ions used to make ionic/organometallic cores complexes include ruthenium(II), osmium(II), iron(II), iridium(II), and various lanthanides. The influence of the dendrimer on the photophysical and electrochemical response of these cores has been studied. Lastly, fluorescent dyes have been employed as cores for their light emission properties. Tremendous interest in fluorescent dye core dendrimers has developed in the field of organic light-emitting diode (OLED) research.

Table 1. Classification of the functional cores used in dendrimer chemistry. Molecule

Properties

Ref.

Porphyrins/Phthalocyanines Si- Phthalocyanine Zn-Porphyrin Pd-Porphyrin Fe-Porphyrin Free-base Porphyrin

U U, E, F, LB P, LB U, E F

[36] [47, 64, 65, 74–76] [80] [72] [50, 63]

2. ENCAPSULATION Encapsulation has been defined as altering an observable physical/chemical property of a molecule by surrounding it with material. The very simplest example of encapsulation is changing the solubility of a core molecule. Although influencing solubility can be accomplished via methods other than encapsulation, in the case of dendrimers, such control of solubility has turned out to be extremely useful, especially in OLED research. Most experiments that study the role of dendritic encapsulation focus on the influence of encapsulation on photophysical behavior, electrochemical behavior, and the kinetics and thermodynamics of ligand binding. Altering these properties is less straightforward than making something soluble and influencing them requires controlling the nanoenvironment around the core [30]. Generally, correlation of a change in one of these three properties, for any given core molecule, to an increase in the generation of a dendrimer indicates encapsulation.

2.1. Effects on Photophysical Properties Observation of photophysical properties has been an extremely useful tool for the evaluation of dendritic encapsulation. The absorbance and/or emission of light by a core is easy to measure, is nondestructive, and is extremely sensitive to the effects of dendritic encapsulation. Fréchet and co-workers first reported dendrimer encapsulation when shifts in the UV-vis absorbance maxima were observed for a p-dimethylamino-nitrobenzene core aryl-ether baseddendrimer (Fig. 2) in a variety of solvents [46]. In all solvents used for UV absorbance, a noticeable increase in the absorbance maximum occurred between the third- and fourth-generation dendrimers. This increase was interpreted

Inorganic Clusters MoV OS4 Fe4 S4 Mo6 Cl8 Re6 Se8

EPR, E E U, E

[37] [39, 66] [42] [40, 41]

Organometallic/Ionic Complexes Ir(ppy)3 Lanthanide, (Ln3+ )–(RCOO− )3 Ferrocene Methyl Viologen

L, U [15] FT-IR, U, E [59, 60] E [67, 68] E [69, 70] Dyes

4-(N-methylamino)-1-nitrobenzene p-dimethoxybenzene Tryptophan Coumarin 343, Pentathiophene Rhodamine B 4-(dicyanomethylene)-2-methyl-6(4-dimethylaminostyrl)-4H-pyrane 3-[dimethylamino]phenoxy group Coumarin 343 Dansyl chloride (S)-1,1′ -bi-2-naphthol

U F F L F F

[46] [48] [49] [51] [52] [81]

F F FQ F, LB

[57] [62] [77] [78]

Note: U = UV-vis absorbance, F = fluorescence, FQ = fluorescence quenching, P = phosphorescence, L = luminescence, E = electrochemistry, LB = ligand binding, EPR = electron paramagnetic resonance, FT-IR = FT-IR.

Figure 2. First Demonstration of Encapsulation—Shifts in UV absorbance maxima were observed between generation 3 and generation 4 of this molecular architecture in various solvents. Adapted with permission from [46], C. J. Hawker et al., J. Am. Chem. Soc. 115, 4375 (1993). © 1993, American Chemical Society.

330

Dendritic Encapsulation

to mean a significant change in the nanoenvironment around the core indicative of encapsulation. Initial evidence for actual “site isolation” was reported by Aida and co-workers for aryl-ether-based dendrimers with Zn-porphyrin cores [47]. In fluorescence studies, quenching was observed between first-generation free-base dendrimers and first-generation zinc porphyrin dendrimers. No quenching was observed in the same experiments with fourth-generation zinc porphyrin dendrimers substituted. Attenuation of quenching was not universal as it was observed that quenching by vitamin K3 , a much smaller molecule, was actually enhanced as the dendrimer grew larger. Moore and co-workers reported shifts in the fluorescence maximum for phenylacetylene dendrimers with a p-dimethoxybenzene core [48]. Expected bathochromic shifts were observed for lower-generation dendrimers while at higher generations, abnormal shifts were observed that were very solvent dependent. From Aida and Moore’s work it was clear that encapsulation was not just a general effect and could possibly be controlled. Smith and co-workers demonstrated a level of control of encapsulation using dendrimers with tryptophan cores. The dendrons were designed to present a hydrogen bonding environment (Fig. 3) [49]. Thus, in solvents that were not hydrogen donors or acceptors, a dendritic effect was observed as a shift in the maximum emission wavelength for tryptophan. The shift was due to hydrogen bonding between the dendron and the tryptophan core. In solvents that were hydrogen

H3C

O

O

CH3 O

O H3C O

CH3 O

O

O

O

O

O

O

NH

R N

R = H, CH3

O HN

O O O

N H

O

H N O

O

O

O O

O

CH3

O

O

O

HN

O

O CH3

O CH 3 O

O O H3C O

O

O O CH3

Figure 3. Controlled Encapsulation—A second-generation dendrimer for which controlled encapsulation was demonstrated. By altering the solvent from a non-hydrogen bonding to a type that was hydrogen bond donating and/or accepting, the emission spectra of the core tryptophan could be shifted [49].

bonding donors or acceptors, the dendritic effect was not observed due to competitive binding between the solvent and the dendron. Vinogradov and Wilson also demonstrated control of encapsulation through pH manipulation of a zinc porphyrin core polyglutamate dendrimer [50]. A substantial decrease in the fluorescence emission for the porphyrin core was reported as the pH of the solution was decreased. The shift in the periphery from anionic to neutral eliminated electrostatic shielding on the surface of the dendrimer and allowed the porphyrin cores to behave more like free porphyrins. Substantial amounts of self-quenching of the porphyrin cores occurred at lower pH values. Encapsulation and site isolation by dendrimers have been used with great effect in the field of organic light-emitting diode (OLED) research. After an initial illustration by Thompson and Fréchet and co-workers [51], several groups have capitalized on the use of dendrimers to isolate chromophores and prevent exciton quenching which reduces efficiency. Otomo and co-workers synthesized Fréchet-type dendrimers with rhodamine B cores [52]. Dendrimer thin films spun onto silica, 10% by weight of dendrimer, had no loss in fluorescence emission intensity and demonstrated mirrorless laser emission. Site isolation and laser emission were also observed in a dendrimer host/guest system with reported nonlinear increases in emission in response to excitation [53]. Following several previous illustrations [54–56], Markham et al. most recently illustrated green phosphorescence in LEDs made from fac-tris-2phenylpyridine iridium, Ir(ppy)3 , core dendrimers [15]. In addition to increased luminescence efficiency due to site isolation of the core, the dendrimers altered the solubility of the Ir(ppy)3 core. This allowed the Ir(ppy)3 normally evaporated to form an OLED, to be spin-cast in thin layers up to 20% by weight. Dendrimer encapsulation also can have a pronounced energy-harvesting “antenna” effect on photoactive cores. Stewart and Fox [57] and Devadoss et al. [58] first illustrated that energy could be collected from chromophores in the periphery of the dendrimer and funneled the core of the molecule. Kawa and Fréchet demonstrated that this effect correlated to increasing generation in lanthanide ion/tricarboxylate core dendrimers (Fig. 4) [59, 60]. Enhanced luminescence could be observed by (a) changing the lanthanide ion used to form the dendrimer complexes, (b) site isolation of the core by the dendrimer blocking self quenching, and (c) increasing dendrimer generation leading to enhanced energy collection and transfer to the core. Fréchet and co-workers have also used the polyvalent nature of dendrimers to enhance the generation-dependent “antenna” effect by attaching light-harvesting chromophores to the periphery of the dendrimer. Energy transfer from the periphery chromophores to the core dye was highly efficient [61, 62]. The “antenna” effect appears to be influenced greatly by dendrimer conformation. Jiang and Aida reported fully substituted dendritic porphyrins had much greater quantum yields for energy transfer than the partially substituted analogues [63]. Fluorescence emission from the series of porphyrin dendrimers was nonlinear in response to the number of dendritic substituents.

Dendritic Encapsulation

Figure 4. Dual-Purpose Encapsulation—A fourth-generation dendrimer that achieves two effects. The aromatic rings serve as lightharvesting units and funnel the energy to the lanthanide core resulting in emission. This emission was enhanced due to the steric shielding provided by the dendrimer which prevents self-quenching by the lanthanide cores. Reprinted with permission from [60], M. Kawa and J. M. J. Fréchet, Chem. Mat. 10, 286 (1998). © 1998, American Chemical Society.

2.2. Effects on Electrochemical Properties Electrochemically active cores have also been instrumental for investigating dendritic encapsulation [9, 31]. Electrochemically active cores have proven useful because two properties, the rate of electron transfer and thermodynamic redox potential, can be altered by encapsulation. Attenuation of electron transfer rates is primarily influenced by steric hindrance and is distance dependent. However, the type of units within the dendrons also has an effect. Encapsulation effects on redox potential, in contrast to electron transfer attenuation, have been less well understood, though some progress has been made [13]. Multiple dendrimer systems with different types of cores and architectures have been reported in which electron transfer rates were attenuated. Diederich and co-workers reported the first attenuation of electron transfer rates for porphyrin core dendrimers [64]. This attenuation increased as the dendrimer generation increased. Fréchet, Abruña, and co-workers demonstrated electrochemical site isolation for Zn-porphyrin dendrimers (Fig. 5) [65]. Thirdgeneration and larger dendrimers attenuated the rate of electron transfer such that it could not be observed using cyclic voltammetry. Gorman and co-workers elucidated structure–property relationships for electron attenuation using Fe4 S4 cores with two different dendrimer architectures (Fig. 6) [39]. These

331

Figure 5. Site Isolation of a Zn-Porphyrin Core—A zinc–porphyrin core was encapsulated by an aryl-ether-based dendrimer. Electrochemical characterization of the core showed a considerable attenuation of the kinetic electron transfer rate upon going from the secondgeneration to the third-generation dendrimer. The rate of electron transfer was attenuated past that observable by cyclic voltammetry by the fourth generation. Reprinted with permission from [65], K. W. Pollak et al., Chem. Mat. 10, 30 (1998). © 1998, American Chemical Society.

dendrimer architectures exhibited remarkable, and seemingly contradictory, behaviors in solution and in thin films. In N,N-dimethylformamide (DMF) solution, attenuation of the heterogeneous electron transfer rates was observed from the Fe4 S4 cores to a Pt electrode for both “flexible” and “rigid” dendrimers as generation increased (Fig. 6). The “rigid” architecture was more effective at encapsulating the core than the “flexible” architecture. Correlating this behavior with the results of computational conformational searches, it was suggested that the distance of electron transfer in the “rigid” architecture was greater than that of a “flexible” architecture for a given molecular weight as the former held the core more towards the center of the molecule than the latter. When the electron transfer rates were correlated to an approximation of the average distance of electron transfer, the “flexible” architecture attenuated the rate more than the “rigid” architecture for that given, approximate distance. In these experiments, in solution, very little shift was observed in the thermodynamic redox potential of the Fe4 S4 core with increasing dendrimer generation. However, when the same series of “flexible” dendrimers were spin-cast onto a platinum electrode surface, the thermodynamic redox potential shifted dramatically with increasing dendrimer generation [66]. It was proposed that, in solution, solvent could penetrate the dendrimer and define the microenvironment around the iron-sulfur cluster. In the film,

332

Figure 6. Effects of Repeat Unit on Encapsulation—“Flexible” and “rigid” Fe4 S4 core dendrimers were studied to determine the effect of repeat unit on conformation and subsequent electron transfer rate attenuation. Generation 3 dendrimers are shown (where the circled-D represent arms identical to the one drawn). Analogous generation one through four dendrimers were studied. Reprinted with permission from [39], C. B. Gorman et al., J. Am. Chem. Soc. 121, 9958 (1999). © 1999, American Chemical Society.

however, solvent was largely excluded, and the differing size of dendrimers provided a difference in microenvironment at different generations. As the generation of the dendrimer increased, the effective microenvironment around the ironsulfur cluster became more hydrophobic, making the cluster thermodynamically more difficult to reduce. Kaifer and co-workers have demonstrated controlled encapsulation using ferrocene core single-arm offset dendrimers (Fig. 7) [67]. By altering the pH of the solution in which electrochemistry was performed, it was proposed that the carboxylate periphery could be made to orient towards an amine-SAM covered electrode surface, effectively stopping free rotation of the molecule. This net effect was detected as attenuation of electron transfer of the ferrocene core which had been correspondingly oriented away from the electrode surface. In further experiments with similar ferrocene core single-arm offset dendrimers, attenuation of electron transfer in both organic and aqueous solvents was demonstrated [68].

Dendritic Encapsulation

While these experiments provide a good basis to rationalize the factors for altering electron transfer via encapsulation, some recent results do not fit the simple pattern that increasing size results in increasing encapsulation. Kaifer and co-workers and Balzani and co-workers independently synthesized identical methylviologen core dendrimers up to the third generation (Fig. 8) [69, 70]. Both groups found no correlation between the electron transfer kinetics and dendrimer generation. The simplest explanation for this behavior would be that the dendrimers were not sufficiently large to encapsulate the core. However, Alvarro et al. worked with the same molecules and reported that electron transfer quenching with anthracene was inhibited by the second-generation dendrimer [71]. The interpretation of this data with regards to encapsulation remains unresolved. The other aspect of electrochemistry influenced by encapsulation is the thermodynamic redox potential. Fréchet, Abruña, and co-workers have observed shifts in redox potential in Zn-porphyrin core dendrimers [65]. Diederich and co-workers reported solvent-dependent effects for various Fe-porphyrin dendrimers [64, 72]. Stone and co-workers reported similar effects for ferrocene core dendrimers [73]. Gorman and co-workers, however, reported no effect on the redox potential of Fe4 S4 cores for two different dendrimer architectures in solution [39]. The same dendrimers in thin films, however, exhibited tremendous changes in redox potential [66]. A model in which the difference in charge states between the oxidized and reduced forms of a core and the relative hydrophobicity of a dendrimer has been advanced to rationalize the results [13]. However, it is unclear how generally applicable this model will be.

2.3. Effects on Ligand Binding The effects of encapsulation on ligand binding to core molecules have been exhibited in diverse dendrimer–core systems. The principal influence is primarily steric in nature. The type of pockets that a dendrimer architecture establishes around the core is not well understood. Aida and co-workers demonstrated inhibition binding of a dendritic imidazole to a Zn-porphyrin core dendrimer [74]. Suslick and co-workers reported shape selective ligand binding in a series of Zn-porphyrin dendrimers [75, 76]. Pockets are built around the Zn-porphyrins with dendrons, limiting, but not completely eliminating access. Kaifer and co-workers also demonstrated hindrance of cyclodextrin binding for their ferrocene-core offset dendrimers [68, 77]. Pugh et al. reported the ability to discriminate chiral compounds with fluorescent dendrimers [78]. Vinogradov and co-workers showed control of O2 binding to a Pd-porphyrin core through prudent selection of the encapsulating dendrimer [79]. In this last example, of three different watersoluble dendrimers synthesized, only the Fréchet-style aryl ether dendrimer was capable of inhibiting O2 quenching of the porphyrin fluorescence emission (Fig. 9). The other two dendrimer architectures permitted O2 access to the core.

333

Dendritic Encapsulation

Figure 7. Controlled Encapsulation with an “On-Off” Effect—By controlling the pH of solution, electron transfer to and from the ferrocene core could be controlled. The dendrimer was proposed to control the orientation of the ferrocene with respect to the electrode. Plots A though D are Osteryoung Square Wave Voltammograms obtained at pHs 4.2, 5.3, 6.0, and 7.0, respectively. Reprinted with permission from [67], Y. Wang et al., J. Am. Chem. Soc. 121, 9756 (1999). © 1999, American Chemical Society.

O O O O O

O

O PF6

O

O

O

O N O

O O

O O

O

N

O O

O

PF6

O

O

O

O

O O O O

Figure 8. Unusual Methyl Viologen Core Dendrimer—A 4,4′ -bipyridinium (viologen) core dendrimer which exhibited no apparent electron transfer rate attenuation in a series of molecules up to the size shown.

334

Dendritic Encapsulation D

D

D

D N

N Pd

N

N

D

D

D

D COOH

COOH HOOC COOH NH

O

D1 =

O

H N

O

O

N H

D2 =

COOH O

O

ACKNOWLEDGMENTS This work was supported by the National Science Foundation Grant CHE-9900072 and the Office of Naval Research Grant N00014-00-1-0633. CSC is the recipient of a GAANN fellowship in Electronic Materials.

COOH O O

D3 =

N H O

O N H

Dendrimer A macromolecule consisting of a core moiety surrounded by one or more arms known as dendrons. Dendron A hyperbranched polymeric arms that form the bulk material of a dendrimer. They are attached to and radiate from the core moiety. Divergent A method of chemical synthesis that begins at the center point of a molecule and builds outwards. Encapsulation The process of influencing access by surrounding with a material. Generation A reference to the number of layers of a dendrimer. A layer is defined by the number of branch points found in the dendrons between the core and the periphery of a dendrimer. Thus, a “generation 4 dendrimer” would have dendrons with four branch points. Monodisperse A collection of objects that all have the same size. With regards to dendrimers, all molecules have the same molecular formula and molecular weight. Nanoenvironment The immediate surroundings around a moiety defined by molecular structure.

O

O N H H O N

O O O

COOH

3

COOH 3 COOH

3

Figure 9. Differential Encapsulation—Three series of water-soluble palladium porphyrin core dendrimers with units illustrated by D1, D2, and D3, respectively. Only the aryl-ether-based dendrimers (repeat unit D1) could prevent quenching of emission from the core by O2 . Reprinted with permission from [79], V. Rozhkov et al., Macromolecules 35, 1991 (2002). © 2002, American Chemical Society.

3. CONCLUSIONS Encapsulation has become a significant role that has been studied for dendrimers. A number of architectures have been explored. Several significant phenomena have been illustrated. In the most general sense, the role of the dendrimer in surrounding and encapsulating the core has been rationalized with simple models. However, in each case discussed above, there are examples that do not neatly fit a simple paradigm. Future work to enlarge the number and type of dendrimer architectures studied will facilitate extension of these models and establishment of more general paradigms for dendritic encapsulation.

GLOSSARY Chromophore Any chemical moiety that absorbs and/or emits light. Convergent A method of chemical synthesis that begins at the edge point of a molecule and builds inwards.

REFERENCES 1. G. R. Newkome, Z. Q. Yao, G. R. Baker, and V. K. Gupta, J. Org. Chem. 50, 2003 (1985). 2. D. A. Tomalia, H. Baker, J. Dewald, M. Hall, G. Kallos, S. Martin, J. Roeck, J. Ryder, and P. Smith, Polym. J. 17, 117 (1985). 3. S. M. Grayson and J. M. J. Fréchet, Chem. Rev. 101, 3819 (2001). 4. C. J. Hawker and J. M. J. Fréchet, J. Am. Chem. Soc. 112 (21), 7638 (1990). 5. C. B. Gorman, Adv. Mater. 10, 295 (1998). 6. L. J. Twyman, A. S. H. King, and I. K. Martin, Chem. Soc. Rev. 31, 69 (2002). 7. R. Kircheis, L. Wightman, and E. Wagner, Adv. Drug Deliv. Rev. 53, 341 (2001). 8. D. Astruc and F. Chardac, Chem. Rev. 101, 2991 (2001). 9. C. M. Cardona, S. Mendoza, and A. E. Kaifer, Chem. Soc. Rev. 29, 37 (2000). 10. G. R. Newkome, E.-F. He, and C. N. Moorefield, Chem. Rev. 99, 1689 (1999). 11. A. W. Bosman, H. M. Janssen, and E. W. Meijer, Chem. Rev. 99, 1665 (1999). 12. M. Venturi, A. Credi, and V. Balzani, Coord. Chem. Rev. 186, 233 (1999). 13. C. B. Gorman and J. C. Smith, Acc. Chem. Res. 34, 60 (2001). 14. E. M. Harth, S. Hecht, B. Helms, E. E. Malmstrom, J. M. J. Fréchet, and C. J. Hawker, J. Am. Chem. Soc. 124, 3926 (2002). 15. J. P. J. Markham, S. C. Lo, S. W. Magennis, P. L. Burn, and I. D. W. Samuel, Appl. Phys. Lett. 80, 2645 (2002). 16. G. M. Dykes, J. Chem. Technol. Biotechnol. 76, 903 (2001). 17. H. Kawaguchi, Prog. Polym. Sci. 25, 1171 (2000). 18. S. C. Zimmerman and L. J. Lawless, Top. Curr. Chem. 217, 95 (2001). 19. M. Fischer and F. Vögtle, Angew. Chem.-Int. Edit. 38, 885 (1999). 20. T. Emrick and J. M. J. Fréchet, Curr. Opin. Colloid Interface Sci. 4, 15 (1999).

Dendritic Encapsulation 21. V. Percec, C. H. Ahn, G. Ungar, D. J. P. Yeardley, M. Möller, and S. S. Sheiko, Nature 391, 161 (1998). 22. S. D. Hudson, H. T. Jung, V. Percec, W.-D. Cho, G. Johansson, G. Ungar, and V. S. K. Balagurusamy, Science 278, 449 (1997). 23. V. Percec, P.-W. Chu, G. Ungar, and J.-P. Zhou, J. Am. Chem. Soc. 117, 11441 (1995). 24. R. Esfand and D. A. Tomalia, Drug Discov. Today 6, 427 (2001). 25. E. Tomlinson and A. P. Rolland, J. Control. Release 39, 357 (1996). 26. C. Z. Chen and S. L. Cooper, Adv. Mater. 12, 843 (2000). 27. M. J. Liu and J. M. J. Fréchet, Pharm. Sci. Technol. Today 2, 393 (1999). 28. K. Uhrich, Trends Polym. Sci. 5, 388 (1997). 29. R. M. Kim, M. Manna, S. M. Hutchins, P. R. Griffin, N. A. Yates, A. M. Bernick, and K. T. Chapman, Proc. Natl. Acad. Sci. USA 93, 10012 (1996). 30. S. Hecht and J. M. J. Fréchet, Angew. Chem.-Int. Edit. 40, 74 (2001). 31. C. S. Cameron and C. B. Gorman, Adv. Funct. Mater. 12, 17 (2002). 32. C. B. Gorman, J. C. Smith, R. Sachdeva, W. Y. Su, and H.-W. Jiang, Macromol. Symp. 156, 61 (2000). 33. F. Diederich and B. Felber, Proc. Natl. Acad. Sci. USA 99, 4778 (2002). 34. C. J. Zhong and M. M. Maye, Adv. Mater. 13, 1507 (2001). 35. F. Diederich, Chimia 55, 821 (2001). 36. N. B. McKeown, Adv. Mater. 11, 67 (1999). 37. S. Mondal and P. Basu, Inorg. Chem. 40, 192 (2001). 38. M. J. Hannon, P. C. Mayers, and P. C. Taylor, J. Chem. Soc.-Perkin Trans. 1, 1881 (2000). 39. C. B. Gorman, J. C. Smith, M. W. Hager, B. L. Parkhurst, H. Sierzputowska-Gracz, and C. A. Haney, J. Am. Chem. Soc. 121, 9958 (1999). 40. B. K. Roland, C. Carter, and Z.-P. Zheng, J. Am. Chem. Soc. 124, 6234 (2002). 41. R. Y. Wang and Z.-P. Zheng, J. Am. Chem. Soc. 121, 3549 (1999). 42. C. B. Gorman, W. Y. Su, H. W. Jiang, C. M. Watson, and P. Boyle, Chem. Commun. 10, 877 (1999). 43. R. M. Crooks, B. I. Lemon, L. Sun, L. K. Yeung, and M. Q. Zhao, Top. Curr. Chem. 212, 81 (2001). 44. V. Balzani, P. Ceroni, A. Juris, M. Venturi, S. Campagna, F. Puntoriero, and S. Serroni, Coord. Chem. Rev. 219, 545 (2001). 45. S. Serroni, S. Campagna, F. Puntoriero, C. Di Pietro, N. D. McClenaghan, and F. Loiseau, Chem. Soc. Rev. 30, 367 (2001). 46. C. J. Hawker, K. L. Wooley, and J. M. J. Fréchet, J. Am. Chem. Soc. 115, 4375 (1993). 47. R. H. Jin, T. Aida, and S. Inoue, J. Chem. Soc.-Chem. Commun. 16, 1260 (1993). 48. C. Devadoss, P. Bharathi, and J. S. Moore, Angew. Chem.-Int. Edit. Engl. 36, 1633 (1997). 49. S. Koenig, L. Müller, and D. K. Smith, Chem.-Eur. J. 7, 979 (2001). 50. S. A. Vinogradov and D. F. Wilson, Chem.-Eur. J. 6, 2456 (2000). 51. A. W. Freeman, S. C. Koene, P. R. L. Malenfant, M. E. Thompson, and J. M. J. Fréchet, J. Am. Chem. Soc. 122, 12385 (2000). 52. A. Otomo, S. Yokoyama, T. Nakahama, and S. Mashiko, Appl. Phys. Lett. 77, 3881 (2000).

335 53. S. Yokoyama, A. Otomo, and S. Mashiko, Appl. Phys. Lett. 80, 7 (2002). 54. M. Halim, J. N. G. Pillow, I. D. W. Samuel, and P. L. Burn, Adv. Mater. 11, 371 (1999). 55. J. M. Lupton, I. D. W. Samuel, M. J. Frampton, R. Beavington, and P. L. Burn, Adv. Funct. Mater. 11, 287 (2001). 56. J. M. Lupton, I. D. W. Samuel, P. L. Burn, and S. Mukamel, J. Chem. Phys. 116, 455 (2002). 57. G. M. Stewart and M. A. Fox, J. Am. Chem. Soc. 118, 4354 (1996). 58. C. Devadoss, P. Bharathi, and J. S. Moore, J. Am. Chem. Soc. 118, 9635 (1996). 59. M. Kawa and J. M. J. Fréchet, Thin Solid Films 331, 259 (1998). 60. M. Kawa and J. M. J. Fréchet, Chem. Mat. 10, 286 (1998). 61. A. Adronov and J. M. J. Fréchet, Chem. Commun. 18, 1701 (2000). 62. A. Adronov, S. L. Gilat, J. M. J. Fréchet, K. Ohta, F. V. R. Neuwahl, and G. R. Fleming, J. Am. Chem. Soc. 122, 1175 (2000). 63. D. L. Jiang and T. Aida, J. Am. Chem. Soc. 120, 10895 (1998). 64. P. J. Dandliker, F. Diederich, M. Gross, C. B. Knobler, A. Louati, and E. M. Sanford, Angew. Chem.-Int. Edit. Engl. 33, 1739 (1994). 65. K. W. Pollak, J. W. Leon, J. M. J. Fréchet, M. Maskus, and H. D. Abruña, Chem. Mat. 10, 30 (1998). 66. C. B. Gorman and J. C. Smith, J. Am. Chem. Soc. 122, 9342 (2000). 67. Y. Wang, C. M. Cardona, and A. E. Kaifer, J. Am. Chem. Soc. 121, 9756 (1999). 68. C. M. Cardona, T. D. McCarley, and A. E. Kaifer, J. Org. Chem. 65, 1857 (2000). 69. R. Toba, J. M. Quintela, C. Peinador, E. Román, and A. E. Kaifer, Chem. Commun. 9, 857 (2001). 70. P. Ceroni, V. Vincinelli, M. Maestri, V. Balzani, W. M. Müller, U. Müller, U. Hahn, F. Osswald, and F. Vögtle, New. J. Chem. 25, 989 (2001). 71. M. Alvaro, B. Ferrer, and H. Garcia, Chem. Phys. Lett. 351, 374 (2002). 72. P. Weyermann, F. Diederich, J. P. Gisselbrecht, C. Boudon, and M. Gross, Helv. Chim. Acta 85, 571 (2002). 73. D. L. Stone, D. K. Smith, and P. T. McGrail, J. Am. Chem. Soc. 124, 856 (2002). 74. Y. Tomoyose, D. L. Jiang, R. H. Jin, T. Aida, T. Yamashita, K. Horie, E. Yashima, and Y. Okamoto, Macromolecules 29, 5236 (1996). 75. P. Bhyrappa, G. Vaijayanthimala, and K. S. Suslick, J. Am. Chem. Soc. 121, 262 (1999). 76. J. Sen and K. S. Suslick, J. Am. Chem. Soc. 122, 11565 (2000). 77. C. M. Cardona, J. Alvarez, A. E. Kaifer, T. D. McCarley, S. Pandey, G. A. Baker, N. J. Bonzagni, and F. V. Bright, J. Am. Chem. Soc. 122, 6139 (2000). 78. V. J. Pugh, Q. S. Hu, and L. Pu, Angew. Chem.-Int. Edit. 39, 3638 (2000). 79. V. Rozhkov, D. Wilson, and S. Vinogradov, Macromolecules 35, 1991 (2002). 80. S. A. Vinogradov, L. W. Lo, and D. F. Wilson, Chem.-Eur. J. 5, 1338 (1999). 81. S. Yokoyama, T. Nakahama, A. Otomo, and S. Mashiko, Colloid Surf. A—Physicochem. Eng. Asp. 198, 433 (2002).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Diamond Nanocrystals Tarun Sharda Seki Technotron Corporation, Tokyo, Japan

Somnath Bhattacharyya Indian Institute of Technology, Kanpur, India

CONTENTS 1. Introduction 2. Deposition Routes 3. Transport Properties 4. Field Emission 5. Optical Properties 6. Mechanical and Tribological Properties 7. Applications 8. Conclusions Glossary References

1. INTRODUCTION “Diamond,” first mined in India thousands of years ago, has been known as the “king of gemstones” by the public for its beauty for centuries. It probably will not be long until in the 21st century the same diamond with a different morphological structure will be known as the “king of materials” to engineers and scientists for its extreme properties with their unique combinations and practical use in manufacturing in the form of films and coatings [1, 2]. The extreme material properties of diamond with some potential applications from their unique combinations can be found in some review articles (for example, see K. V. Ravi in [1] and P. Chalker and S. Lande in [2]), and in addition some new areas of applications such as surface acoustic wave (SAW) devices [2, 3], and micro/nanoelectromechanical systems (MEMS and NEMS, respectively) [4–6] have started emerging. In brief, diamond is very hard and least compressible, is optically transparent from the deep ultraviolet (UV) to the far infrared, has the highest known thermal conductivity of all materials at room temperature, has low thermal expansion coefficient at room temperature (08 × 10−6 K) that is comparable to that of invar, exhibits low or “negative” electron affinity, and is biologically compatible and very resistant to chemical ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

corrosion [1, 2, 7, 8]. Diamond is much more resistant to X-rays,  radiation, UV light, and nuclear radiation than the other semiconductors such as Si and Ge. Also, it is a very good electrical insulator (with room temperature resistivity of ∼1016  cm). However, its resistivity can be changed over the range 10–106  cm by doping, thus becoming a semiconductor with a wide bandgap of 5.5 eV. The great interest in diamond also comes from its applications in electronic devices as it has very high mobility (about 2000 cm2 /V s), breakdown field (10 MV/cm), and saturation velocity (about 25 × 107 cm/s) compared to other semiconductors [8]. At the same time, there are other diamond related carbon materials that have properties in between those of diamond and graphite. The properties of these hard carbon materials vary, depending on the concentration of sp3 -(diamond) and sp2 -(graphite) bonded carbon in the films. A classification of diamond and related materials in terms of their structure, resistivity, and energy gap is given in Table 1. The energy difference between thermodynamically stable graphite and metastable diamond is 0.9 kcal/mole. Nanodiamond is a metastable material with intermediate energy, close to ta-C having sp3 bonds more than 90% [9, 10]. Before progressing further let us discuss diamond and related materials in some more detail.

1.1. Synthetic Diamond Scarcity and cost of the natural diamond led scientists to find means of making diamond in the laboratory. Although there have been several attempts to synthesize diamond from various sources of carbon since it was discovered to be an allotrope of carbon, two main techniques were invented, both in the 1950s; high pressure and high temperature (HPHT) synthesis [1, 7] and chemical vapor deposition (CVD) synthesis [1, 2]. To date, since its discovery, most of the demands to supply diamond for industrial uses, primarily for material cutting, grinding, and polishing, are met by the HPHT technique. In this technique graphite is subjected to tens of thousands of atmosphere and heated to over 2000 K in the presence of some metal catalyst crystallizing it in diamond. However, this process is intrinsically

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (337–370)

Diamond Nanocrystals

338 Table 1. Properties of diamond, graphite, and amorphous carbon.

Material Diamond Nanocrystalline diamond Tetrahedral amorphous carbon Amorphous carbon Graphite

Structure

Resistivity ( cm)

Energy gap (eV)

100% sp3 -C >90% sp3 -C

>1010 >106

5.5 2–5.5

>80% sp3 -C

∼104

∼2

∼10–50% sp3 -C 0% sp3 -C

∼10−1 10−4

a few meV 0

limited in its ability to cover surfaces in the form of thin film coatings and to produce large sizes. These limitation are overcome by another technique (i.e., CVD synthesis of diamond) in which diamond is grown at subatmospheric pressures in the form of thin films [2]. Although the invention of this technique dates back to 1950s, its early impracticalities in terms of economical viability due to extremely low growth rate and co-deposition of graphitic phase of carbon did not allow it to develop [11]. In late 1960s to early 1980s it was discovered that presence of atomic hydrogen can preferentially etch the co-depositing graphitic phase and can improve the growth rate significantly [12, 13]. The current surge in CVD diamond research can be traced back to 1982 when the Japanese group at the National Institute for Research in Inorganic Materials (now the National Institute of Advanced Industrial Science and Technology) published details of their work clearly indicating the ways and means of generating atomic hydrogen in the growth chamber to grow good quality diamond on nondiamond substrates at significant rates (∼1 m/hr) [14]. Following this the same group came up with another technique in 1983 and this stimulated the worldwide interest in CVD diamond [15]. In general, in this technique, diamond can be nucleated on nondiamond substrates and an oriented or polycrystalline thin film can be deposited by dissociating the precursor gases at low pressures. These polycrystalline films show the same characteristics as natural and synthetic diamond and have high potential for various applications [1, 2, 8]. Generation of atomic hydrogen and the excitation of hydrocarbons can be achieved by various means and can broadly be classified as thermal CVD and plasma CVD. In thermal CVD processes, decomposition of precursor gases is accomplished by thermal activation whereas in plasmas it occurs by electron–molecule reactions. In high temperature plasma torches both mechanisms can be important. Thermal activated CVD includes hot filament and oxyacetylene torches. Plasma CVD includes low and medium pressure microwave (2.45 and 0.915 GHz) and dc plasma, dc plasma jet, low pressure radio frequency (rf) plasma, etc. [2].

applications. For example, diamond is well suited for use as a protective optical coating but diamond films with high surface roughness cause attenuation and scattering of the transmitted signals restricting their use in optical coatings. In order to overcome the problem of surface roughness of diamond films either postpolishing is to be adopted or naturally grown smooth films are to be developed without compromising much of their hardness and other useful properties. However, postpolishing is expensive and time consuming [18–20] and it may be better to concentrate on as-grown smooth diamond films [19–51]. One way to reduce the surface roughness during growth is to control crystalline orientation with (100) facets being parallel to the film plane [52], or by reduction of the grain size (i.e., reduction of grain sizes from micrometers to nanometers) [9, 35, 37, 53]. This is the reason that in the last few years efforts in the area of diamond films grown by CVD have diverted from conventional microcrystalline diamond (MCD) films, having a grain size of a few hundred nanometers to several tens of micrometers [9, 35, 53–55], to nanocrystalline diamond (NCD) films, having a grain size from 2 nm to a few hundred nanometers [9, 19–51, 53, 56–62]. Moreover, this change of perception is a result of NCD films not only being smooth but also being more adaptive to the CVD diamond applications and having enhanced electronics and other properties. This is reflected in their cost effective and superior performances in the areas of electron field emission [63–69], optical transparency and protective coatings [21–23, 28, 38–40, 45, 51, 53, 70–72], tribology [19, 33, 34, 36, 37, 48, 73–75], SAW devices [3], and MEMS [5, 6]. Some of these properties of NCD films and a comparative study with single crystal diamond (SCD)/MCD are listed in Table 2. Moreover, as the size of the grains is changing from micrometer to nanometer, a factor of a million in volume, new properties have started emerging, exploring new areas of applications [76], for example, in X-ray optics, X-ray physics, particle physics, etc. Some important properties of nanocrystalline diamond films are • large surface area of diamond and high density of grain boundaries • drastic increase in surface smoothness relative to microcrystalline diamond • presence of more defects, however, less disorder in the interface • better electron emission than conventional diamond films • less highly oriented grains • important grain boundary conduction • enhanced optical transmission.

1.2. Nanocrystalline Diamond

1.3. Diamond-like Carbon and Tetrahedral Amorphous Carbon

The high surface roughness of the conventionally grown polycrystalline CVD diamond films prevents their immediate use in spite of having outstanding properties with their unique combinations suitable for many applications. The high surface roughness is a major roadblock using these CVD diamond films for various applications [16, 17] such as machining and wear, optical, and thermal management

There is another class of hard carbon materials [1, 77, 78], essentially amorphous in nature and consisting of a mixture of sp2 and sp3 carbon, broadly termed “diamondlike carbon” (DLC). A significant amount of sp3 carbon in the films gives them attractive physical and mechanical properties such as high hardness, high wear resistance, low coefficient of friction, chemical inertness, a high optical bandgap, and high

Diamond Nanocrystals

339

Table 2. Some properties of NCD films. Potential applications of NCDc

SCD/MCDa

NCD

Ref.b

5.5 negative 160

2–5.5 negative 3–4

[30, 53, 70] [153] [63]

field emission display

>1010 p-type possible

>106 both p- and n-type possible

[56] [56, 162]

electrochemical electrodes

transparent from UV to far IR 2.41

∼78–84% transmittance beyond 700 nm ∼2.27–2.35

[45, 53]

optical coatings

[53] (@ 830 nm), [201]

IR windows

Mechanical properties 1. Hardness (GPa) 2. Young’s modulus (GPa)

∼100 1054

>90 ∼1000

[5] [5, 6]

3. Wear resistance 4. Coefficient of friction

high 90% sp3 -C 2–100 2–50 1000 10

>800 10

[47, 170] [3]

SAW devices bioelectronic sensing System

biocompatible

biocompatible, highly selective, and stable

[232]

hardness, thermal conductivity, breakdown voltage, etc. (for details see [1, 2, 7, 8])

smoothness, structural defects, high density of GB, conductivity, hardness, strong adhesion, large stress

this work, [9, 56, 76]

Properties Electronic properties 1. Bandgap (eV) 2. Electron affinity 3. Onset of field emission (V/m) 4. Resistivity ( cm) 5. Doping Optical properties 1. Transmission 2. Refractive index (@ 633)

Structural properties 1. Bonding character 2. Grain size (nm) 3. Surface roughness (nm) 4. Hydrogen content (%) Other properties 1. Thermal conductivity (W/cm K) 2. Thermal stability ( C) 3. SAW phase velocity (km/s) 4. Biological properties Unique characteristics

a b c

electronic devices

wear resistant, protective coatings MEMS/NEMS

Data for SCD/MCD are taken mostly from [1, 2, 7, 8]. References for NCD. Potential applications are a result of combinations of various properties in most of the cases.

electrical resistivity. These properties in the DLC films vary with the concentration of sp2 and sp3 carbon in the films. It has been shown that the properties of DLC films, such as hardness, optical transparency, etc., approach the value of diamond with high concentration of sp3 carbon [79, 80]. The DLC films with high sp3 concentration are termed tetrahedral carbon (ta-C) or amorphous diamond (a-D). The term DLC is mostly used to designate the hydrogenated amorphous carbon (a-C:H) whereas the term ta-C is mostly used to designate the nonhydrogenated amorphous carbon (a-C) [78]. The a-C:H films may contain 50% sp3 carbon whereas it may be as high as 85% in ta-C films. Pulsed laser deposition [72–74, 79, 81, 82], cathodic arc deposition [80, 83], and ion deposition methods [84] have been employed to grow this new class of carbon material. Considering the smooth surfaces having properties close to diamond, NCD and ta-C have emerged from their respective branches of CVD diamond and DLC, as described, as

the materials for intense current and future activities. As mentioned, it is not only because these films are smooth but also because they are more adaptive to the CVD diamond applications, having enhanced electronics and other properties. This makes the basis of this chapter in which some properties of NCD are discussed, emphasizing, in particular, transport and electronic properties, and are compared with the properties of ta-C. First, various deposition routes of NCD are discussed. Second, transport and electronic properties of NCD will be discussed in detail followed by a discussion on optical and mechanical properties. At the end, various other emerging aspects of NCD are also presented.

2. DEPOSITION ROUTES A common feature of the majority of the deposition techniques of CVD diamond films is a high concentration of hydrogen gas (H2 used as one of the constituents with some

340 hydrocarbon gas such as CH4 . A concentration of H2 in excess of 99% (of the gas mixture) is required for the growth of high quality MCD films. The high concentration of H2 results in the generation of a large flux of atomic hydrogen, which is generally believed to play a central role in the various diamond CVD processes. The growth of diamond is said to take place mostly via surface processes of addition and abstraction of radicals from the gas phase [8, 85–87]. Initially, after the nucleation of diamond, the growing diamond surface is to be protected from reconstruction into the graphitelike form. Therefore a high concentration of H is needed to satisfy the dangling bonds left unsatisfied on the growing diamond surface. Subsequently, vacancies are created by hydrogen atoms, which interact with the bonded hydrogen and remove it by converting it into H2 . Methyl radicals (CH3 , also generated by gas-phase hydrogen abstraction by H, are considered as diamond precursors are then placed on the vacancies in order to promote growth. In addition, atomic H preferentially etches the graphitic phase co-depositing with diamond. Generally, before realizing the importance of growing naturally smooth surfaces, the aim in the area of CVD diamond was to maximize the crystalline quality of CVD diamond. However, diamond grown under nonoptimum conditions, such as lower hydrogen concentration or higher carbon activity in the plasma, gives films with small grain size (e.g., several nanometers). Also, ions are not considered to be good for diamond growth as they enhance lattice disorders and also promote graphitic content in the deposit. On the other hand, in the growth of carbon films by hyperthermal species, ions with energy in the range 40–200 eV produce a high concentration of sp3 carbon in the films [83, 88, 89]. In this case the growth is dictated by a subplantation mechanism, which relies on the shallow subsurface implantation of the carbon ions [88]. Thus the two routes of growth of carbon films are totally different (i.e., one relies on surface and the other on subsurface processes). Moreover, the deposits are predominantly crystalline and amorphous respectively in the two processes. Interestingly, it was shown recently that both the routes can be coupled not only to generate diamond nuclei in the process called biased enhanced nucleation but can also be used to grow smooth NCD films [41, 42]. In the already existing ways to grow NCD, a number of techniques and conditions have been employed [9, 26, 27, 36, 37, 42, 44, 45, 48, 49, 57, 76]. As the need for asgrown smooth diamond films has been realized in the last few years, new routes have also emerged that mostly aim to grow smooth NCD or NCD embedded in a ta-C or a-C matrix. Based on the literature, a brief description of these new routes is as follows.

2.1. Hydrogen Deficient Gas Phase As explained, high hydrogen concentration gives high quality MCD films and one route adopted to reduce the crystalline size to obtain NCD films is to reduce the concentration of hydrogen or increase the carbon activity in the gas phase and on the growing surfaces either by increasing the hydrocarbon constituent in the gas phase or by substituting the hydrogen with other gases, partially or completely. The resultant films vary from phase pure NCD to NCD embedded in ta-C or a-C matrix.

Diamond Nanocrystals

2.1.1. Hydrogen Deficient CH4 /Ar, C60 /Ar, CH4 /N2 , CH4 /He One such process that has been studied in detail was developed at Argonne National Laboratory, USA [5, 9, 10, 19, 24, 31, 32, 35, 46, 56]. In this process, carbon dimer (C2 ) is used as a reactive species in hydrogen deficient (CH4 /Ar or C60 /Ar) microwave plasma CVD [9]. C2 is produced by replacing molecular hydrogen by argon and using CH4 or C60 as precursor gases and also by using N2 /CH4 as the reactant gases in a microwave plasma CVD system. The NCD films grown on seeded substrates by this technique are composed of 3–15 nm diamond crystallites with up to 1–10% sp2 carbon residing at the boundaries [9]. Konov et al. [25] also reported growth of smooth NCD films using CH4 /H2 /Ar mixtures with a methane percentage CH4 /(CH4 + H2 ) varied in the range of 10 to 100% in a dc arc plasma deposition system. Their NCD films grown on Si substrates, seeded with ultrafine (5 nm) diamond particles, consisted of 30–50 nm size diamond crystallites. In the partially substituted hydrogen Nistor et al. [29] produced 1–2 m thick NCD films from CH4 /H2 /Ar mixtures by direct current plasma CVD. The Ar concentration was kept constant at 50% and methane concentration CH4 /(CH4 + H2 ) was varied from 3 to 100%. Similar to Konov et al. [25] they also obtained nanodiamond films without even using hydrogen in the feed gas. Smooth NCD films with grain size of 10–50 nm were obtained on substrates seeded with 5 nm diamond powder. Although their transmission electron microscopy (TEM) studies did not reveal amorphous carbon in significant amounts, the nanocrystallites were found to be highly defective with many twins and other planar defects. The grain boundaries were presumably decorated with disordered sp3 -bonded carbon and sp2 -bonded amorphous carbon. Lin et al. [90] studied the CH4 /H2 /Ar process in a hot filament CVD (HFCVD) system and proposed a compositional map demarcating regions for well-faceted diamond growth, NCD growth, and nondiamond deposition. Well-faceted MCD (2–10 m) grow using Ar up to 90%. They report a change in microstructure from MCD to NCD, with a grain size smaller than 50 nm, at 95.5% Ar addition. Yang et al. [91] also used the CH4 /H2 /Ar process to grow NCD films and observed decreasing grain size and that the surface morphology of the films changes gradually from faceted to ballastlike diamond with increasing Ar concentration. The formation of NCD was seen when more than 90% Ar was added in the CH4 /H2 /Ar plasma. In another study [92] a low concentration (10–15%) of nonuniform distribution of nanodiamonds in amorphous carbon was detected at temperatures as low as 15  C from decompositions of CH4 /Ar-rich plasmas generated at 35 kHz. Amaratunga et al. [93, 94] reported mixed phase films containing diamond crystallites (10–200 nm) embedded in a nondiamond carbon matrix while using CH4 /He plasmas. Bi et al. [3] used Ar, H2 , and CH4 in the ratios 100:4:1, respectively, to grow thick NCD films for a SAW device. The root mean square (rms) surface roughness in their case was 50 nm while the mean average grain size, estimated by scanning electron microscopy (SEM), was approximately 30 nm.

Diamond Nanocrystals

2.1.2. CH4 -Rich CH4 /H2 with or without N2 and O2 Yoshikawa et al. [48] prepared smooth NCD films having approximately 10 nm size grains, as observed by field emission SEM, by a microwave plasma CVD (MPCVD) system using a high methane concentration (10 vol%) in hydrogen on substrates pretreated by diamond powder of sizes 5 and 250 nm. Heiman et al. [47] used 9.91% methane in hydrogen gas mixture in a dc glow discharge. Smooth NCD films of 1 m thickness with grain size in the range of 3–5 nm were obtained. Catledge and Vohra [36, 37] reported a new regime for synthesis of nanostructured diamond films on Ti-6Al-4V substrates. They used high pressures (125 Torr) and high methane concentration (5–15%) in balanced hydrogen and nitrogen in a MPCVD system to obtain predominantly diamond nanocrystallites of 13 nm average size, as estimated by X-ray diffraction (XRD) patterns, in a matrix of ta-C. The films are shown to be significantly smooth compared to the conventionally grown MCD films. At these processing conditions the microwave plasma is more highly activated with C2 and atomic hydrogen. In another report they added varying concentrations of nitrogen in the plasma with 15% CH4 and found that the grain size decreases up to the ratio of 0.1 of N2 /CH4 before saturating at higher nitrogen concentrations. The amount of ta-C also increases in the same fashion and, as expected, the surface roughness also follows the same trend. It was suggested that the change in the film morphology is not a result of N2 incorporation in the films but is a result of N2 related surface processes such as microtwinning due to changes in gas-phase chemistry and surface kinetics. The same group, in their another study [95], modeled the gas-phase composition for MPCVD grown diamond with H2 /CH4 /N2 mixtures using a thermodynamic equilibrium package and compared it with their experimental results by growing diamond films using the same mixture with the MPCVD system. A correlation was shown between the calculated composition of CN radical in the gas phase and the experimentally determined nanocrystallinity of the samples grown with different N2 additions in the H2 /CH4 plasma indicating the role of CN radical in the nitrogen-induced diamond nanocrystallinity in MPCVD. Yang et al. [51] grew transparent diamond films with crystallite size below 70 nm in hydrogen and methane microwave plasma CVD on quartz substrates, ultrasonically pretreated by 0.5 m diamond powder for 30 min. They studied surface morphology and optical and mechanical properties of the films by varying methane concentration in hydrogen plasma in the range of 1 to 4% while introducing 0.2% oxygen after the nucleation and keeping the substrate temperature 500  C at 4 kPa pressure and 1.5 kW microwave power constant. Grain size and surface roughness were observed to decrease with methane concentration with a significant reduction in grain size occurring at 3% methane, which further reduced at 4% methane though the nanodiamond concentration in the films also goes up. The crystallite size, as estimated by TEM, near the interface was 30 nm and increases to 65 nm near the growth side. A thickness of 1–1.2 m was obtained in 4 h growth of the films. Chen et al. [38–40, 45] used the MPCVD system to grow NCD films on quartz substrates,

341 ultrasonically pretreated (using 4 nm diamond powder) for 8 h, using CH4 /H2 /O2 plasmas in which CH4 was varied between 1.5 and 42% while keeping O2 constant at 0.1%. The average surface roughness of their 0.5 to 1 m film varied between 6 and 13 nm. They showed that the surface roughness of the NCD films decreases with increasing methane concentration and has a great impact on the optical transmittance of the NCD films. Wu et al. [22] demonstrated growth of NCD (20–100 nm grain size) having a surface roughness of ∼15 nm by low pressure (5 Torr) and low microwave power (450 W). Erz et al. [23] used 10% CH4 in hydrogen to grow 800 nm NCD film. Zarrabian et al. [28] used electron cyclotron resonance (ECR) plasma to grow NCD grains of size 4–30 nm embedded in DLC. Jiang et al. [59] showed that NCD can be grown with MPCVD using 5% CH4 on Si ultrasonically pretreated by diamond or SiC powders. The average grain size estimated by TEM analysis revealed a NCD grain size of 30–40 nm. The substrates, mechanically abraded by Al2 O3 grit, give NCD embedded in a-C matrix at the same conditions. Wu et al. [66, 67] grew NCD films using N2 /CH4 /H2 in a MPCVD system. In the first series the CH4 was varied between 2.1 to 8.4 sccm in 150 sccm of nitrogen without any hydrogen. In the second series hydrogen was introduced and was varied in the range of 0–10 sccm while keeping the methane constant at 3.5 sccm. NCD films with grain size of 8 nm embedded in a-C matrix were obtained without any hydrogen which increases to 20 and 50 nm respectively at 5 and 10 sccm of hydrogen. In another interesting development, composite diamond films with smooth surfaces, was achieved by alternatively depositing conventional MCD and NCD films in a HFCVD reactor [50]. The structure of the layers was such that the bottom and top layers were respectively MCD and NCD. SEM and atomic force microscopy (AFM) analyses of the films revealed 2–3 m granules consisting of 10–100 nm diamond grains unlike the well-faceted micrometer size grains observed in conventional MCD films. The surface roughness was less than 100 nm. These films have dielectric properties similar to conventional MCD films with smooth surfaces and are considered to be promising for MEMS devices.

2.1.3. CO-Rich CO/H2 Mixtures Lee et al. [26, 27] developed a low temperature process (350  C < T < 500  C) for low power but high growth rate (up to 2.5 m/hr) NCD film by MPCVD. Instead of high methane concentration they used CO-rich CO/H2 mixtures and obtained smooth NCD films consisting of 30–40 nm grain size. They observed an interesting growth trend with respect to substrate temperature while using the CO-rich CO/H2 mixture. There is a narrow temperature window, different for different CO/H2 ratios. The temperature at which the peak growth rate is obtained in this window decreases with increasing CO/H2 ratio. For example, the peak of the growth rate is observed to occur at 500  C at a ratio of 8 of CO/H2 that decreases to 450  C at a ratio of 18. Moreover, the growth rate also increases with increasing ratio. Although the growth of NCD was obtained at low temperatures the bulk void fraction also increased significantly compared to the films grown at higher temperatures. Recently,

342 growth of NCD films of size 20 nm was also reported by Teii et al. [57] at low pressure (80 mTorr) at 700  C by inductively coupled plasma employing CO/CH4 /H2 and O2 /CH4 /H2 . In this case a positive biasing of 20 V was applied to the substrate to reduce the influence of ion bombardment. The films consist of ball shape grains of size 100 nm, which are composed of 20 nm NCD grains.

2.2. Biased Enhanced Growth As the reader may have noticed, in the growth route for NCD by hydrogen deficient plasmas, mostly the substrates were pretreated externally before deposition ultrasonically or mechanically using diamond or other abrasive powders. There is another well-established method to nucleate diamond internally in conventional growth of MCD films called biased enhanced nucleation (BEN) [96] in which the substrates are biased negatively in a relatively hydrocarbonrich mixture of hydrocarbon–hydrogen precursor gases. This method, which results in a high density of diamond nucleation (1010 /cm2 or more), is the first step followed by another step or two steps to grow heteroepitaxial diamond films. In order to achieve overgrowth of diamond on these nuclei, the BEN process is followed by conventional growth in which the bias is put off and the growth is continued with lower hydrocarbon percentage in the gas phase. It is found out that in this later stage of growth only the stable parts of the nuclei continue growing with while the remaining getting etched off in the process due to high concentration of hydrogen in the conventional step of growth [97]. To obtain an NCD film, Sharda and Soga [61] considered achieving higher densities of diamond nucleation, similar to the BEN process, followed by their continuous growth in the later stages while maintaining the high density. In order to achieve this they extended the BEN in microwave plasma CVD for the whole growth process and termed their process biased enhanced growth (BEG) in which they obtained diamond nucleation and growth on silicon substrates in a single process [41–43, 76]. They used 5% CH4 in balanced H2 and controlled the bias current density in a special arrangement in a MPCVD system to study the growth in a different combination of conditions (bias current density, bias voltage, and microwave power). A different regime of growth was indeed observed which resulted in NCD films having rms surface roughness of 5–30 nm composed of grains of 2–30 nm with growth rates of 0.5–2 m/hr at 1000 W microwave power and at substrate temperatures in the range of 400–700  C. In a similar development, Gu and Jiang [44] also used biasing to not only nucleate diamond but also to grow films. Their growth process, though, uses continuous biasing to grow NCD but in three steps unlike a single step process for both nucleation and growth by Sharda et al. [41–43, 61]. No biasing and methane were used in the first stage for 40 min at 860  C, which was meant just to remove the oxide layer, followed first by a nucleation stage in which the substrates were biased at −120 V at 5% methane (at 840  C, 20 mbar pressure, and 800 W microwave power) and second by the growth step in which the substrate was biased up to −140 V while decreasing the methane concentration in the range of 0.5 to 5% and changing other conditions as well. It was observed that the grain size decreases from

Diamond Nanocrystals

0.3 m at −100 V to 40 nm at −140 V. The growth rate was 0.25 m/hr and surface roughness, as measured by a surface profilometer, of 1 m thick film was 5 nm at −140 V. There was a marked difference in the adhesion of the films in case of Sharda et al. [62, 76, 98] and Gu and Jiang [44]. The BEG process by Sharda et al. results in a strong adhesion that allows one to grow any thickness, irrespective of amount of the stress in the films, whereas the NCD films grown by continuous biasing by Jiang and Jia [49] cannot be grown beyond 2 m due to relatively weak adhesion as the compressive stress in the films make them delaminate from the substrate. Groning et al. [99] grew NCD films on silicon substrates to study their field emission properties with and without biasing. In case of no biasing they used high substrate temperatures (950 and 980  C) and high methane fraction (5%) and in the case of biasing at −250 V they used a relatively low substrate temperature (850  C) and lower methane fraction (2%). In the latter case the surface morphology was ballastlike with large ball-like structures of 2–3 m in diameter consisting of diamond nanocrystallites of size 10–50 nm. In an effort by Yang et al. [100], the effect of bias in forming NCD in the MPCVD system was studied. The NCD films were grown on scratched substrates unlike the films grown on externally untreated mirrorlike substrates by Sharda et al. or Jiang et al., as described. A continuous dc negative biasing in the range of 0–250 V was applied to the substrates during deposition while using a 1% CH4 /H2 mixture for 3 hr to study crystallinity, morphology, and growth rate of the diamond films. The substrate temperature was maintained at 900  C while using 5.3 kPa chamber pressure and 1000 W microwave power. A change in the morphology with biasing voltage was reported from triangle-faceted grains of 1–3 m size (up to a bias of 50 V) to square-faceted grains of 1–2 m size (at 100 and 150 V). Smooth NCD films with grain size of 20–40 nm, as measured by TEM, were obtained at 250 V. The growth rate was reported to decrease with biasing beyond 50 V with the NCD growth rate being 0.3–0.4 m/hr. It should be noted that the effect of biasing on the growth rate in this study is different than the effect on biasing in the BEG process reported by Sharda et al. [98]. In the later study, the layer thickness in 60 min growth increases from 1 to 2 m with increasing biasing from 200 to 320 V. In another interesting effort by Yang et al. [101] to grow NCD films, the two main methods to grow NCD, hydrogen deficient argon plasma using methane and fullerene (Section 2.1.1) and biased enhanced growth (Section 2.2), were coupled to study their combined effect on the grain size, microstructure, and growth rate of the diamond films in the MPCVD system. The NCD films with grain size of 50–100 nm were grown using an Ar concentration of 90–99 vol% while continuously biasing the substrate at −50 V. This combination gives a growth rate of 0.7–0.8 m/hr of NCD films in the MPCVD system, apparently larger than what the authors obtained by individual processes. However, it should again be mentioned that even the maximum growth rate (0.8 m/hr) of NCD films, with or without Ar addition, reported by Yang et al. [100, 101] is still lower than the maximum growth rate (∼1.9 m/hr) of NCD films by the BEG process reported by Sharda et al. [98].

Diamond Nanocrystals

More studies on the growth of NCD films began appearing, with Zhou et al. [58, 102] reporting growth of nanodiamond a-C composite in a HFCVD system. The continuous biasing process was termed as prolong biased enhanced nucleation in which they obtained NCD (∼10 nm) embedded in a-C matrix at 1% CH4 in hydrogen. They obtained 1 m thick film in 7 h of deposition. The rms surface roughness of their films measured by AFM was ∼20 nm. Nonfaceted granular structures of size 100–200 nm are seen in SEM which consists of ∼10 nm diamond grains embedded in a-C matrix, as seen in low magnification TEM. In another development, growth of smooth nanodiamond films (NCD embedded in a-C matrix) on different substrates was reported by Kundu et al. [30] at room temperature using dc magnetron sputtering of vitreous carbon in an argon and hydrogen (0–10%) plasma in the pressure range of 45–90 mTorr. The average grain size, as examined by TEM, was found to be ∼55 nm on quartz and ∼75 nm on Si and Mo substrates. The film surface consists mostly of sp2 bonded carbon and was removed to see the NCD embedded in the a-C matrix by SEM. The sp3 /sp2 ratio, as estimated by Fourier transform infrared (IR) increased from 4 to 16 with increasing negative bias voltage applied to the substrate. In an interesting advancement in the synthesis of diamond films, a multilayer structure of MCD and NCD was developed using growth with and without biasing [103]. It was shown in this study that size of the diamond grains could be reduced to a few nm by applying a negative bias during the growth of an MCD film and could again be increased once the bias was put off. This way a layered structure of MCD and NCD can be grown. In this study, the grain size of the initial MCD layer, which was grown for 2 hr, had a grain size of 0, and for ferromagnets: b → 0. 3. For a superconducting layer with a higher Tc : b < 0. The latter case is of theoretical interest because then the order parameter near the surface will increase, which will lead to higher critical fields and larger critical temperatures.

κ = 0.42

κ = 0.71

In κ

Figure 3. The dependence of the characteristics of semi-infinite superconductors on the value of the Ginzburg–Landau parameter . Hc3 is the surface superconducting field when the field is parallel to the surface. Modified from [32].

3. TYPE I AND TYPE II SUPERCONDUCTORS 3.1. Bulk Superconductors Bulk superconductors can be separated into two types through their Ginzburg–Landau parameter  =  T / T : √  < 1/ 2 → type I superconductors √  > 1/ 2 → type II superconductors All superconducting elements except niobium are type I superconductors. Niobium and all superconducting alloys and chemical compounds are type II. The high Tc superconductors also belong to the second group. The dependence of the superconducting characteristics on the value of  is shown in Figure 3. For  < 0"42 the superconductor is a type I superconductor. For fields below the thermodynamical critical field Hc the superconductor is in the Meissner state and all flux Ψ

1 Metal

Superconductor

0 x b

Figure 2. Schematic presentation of the spatial dependence of the superconducting order parameter   at the interface between a superconducting and a normal metal. The extrapolation length b is indicated.

is expelled from the sample. At the critical field the magnetic field penetrates the sample, the superconductivity is destroyed, and the sample becomes normal. For 0"42 <  < √ 1/ 2 ≃ 0"71 the superconductor is still a type I superconductor, but now the Meissner state does not change immediately into the normal state with increasing field. At the field Hc flux can penetrate the inner part of the sample, while a layer remains superconducting near the surface of the sample. At the surface critical field Hc3 the surface becomes normal too and the sample is in the normal √ state. In type II superconductors  > 1/ 2 , on the other hand, a fourth possible state exists. In equilibrium, the Meissner state is only observed at applied fields H0 below the first critical field Hc1 . In the region between the first critical field Hc1 and the second critical field Hc2 the magnetic flux is able to penetrate the sample in quantized units of the flux quantum #0 = hc/2e, called vortices. Abrikosov [33] found that these vortices form a triangular lattice inside the superconductor, the so-called Abrikosov vortex lattice. The state is called the Abrikosov vortex state or mixed state. In the region Hc2 < H0 < Hc3 superconductivity only exists at a thin layer near the sample edges, while the inner side of the sample is in the normal state. For bulk type II superconductors the third critical field Hc3 is approximately equal to 1"69Hc2 . For larger fields superconductivity is destroyed, and the entire sample is in the normal state. The critical fields Hc , Hc1 , Hc2 , and Hc3 depend on the temperature. The H − T phase diagrams for type I and type II bulk superconductors are shown in Figure 4. The magnetization behavior as a function of the external magnetic field is also different for both types. This can be seen from Figure 5. The magnetization of a superconductor is defined as 0  = 1 B − H M 4

(20)

591

Doubly Connected Mesoscopic Superconductors H0

H0 Hc3(T) Surface superconductivity

Hc2(T)

Hc(T) Mixed State Meissner State

Meissner State

Tc0

0

T

Hc1(T) Tc0

0

T

(b) Type-II

(a) Type-I

Figure 4. H − T phase diagrams for type I (a) and a type II (b) bulk superconductors.

  is the magnetic induction and H  0 is the where B = H applied magnetic field. At H0 < Hc a type I bulk superconductor is in the Meissner state, and all flux is expelled from the sample:   = 0 and −4 M  = H  0 . At larger fields, the applied H field penetrates into the superconductor, which becomes  = H  0 and M  = 0. Type II superconductors normal: H  =H 0. are in the Meissner state at H0 < Hc1 and −4 M In the mixed state Hc1 < H0 < Hc2 the absolute value of  decreases with increasing field until the magnetization M it vanishes at the second critical field. Also the surface energy is different for type I and type II superconductors (see, for example, [30]). The expulsion of the external field increases the energy of a superconductor by H 2 /8 per unit volume. It is energetically more favorable that the volume will be divided up into alternate normal and superconducting regions. The creation of such normal regions requires a negative interface surface energy 'ns whose magnitude is such that its contribution exceeds the gain in magnetic energy. The two types of superconductors have a different behavior with respect to the surface energy: Type I ( 'ns > 0) Type II ( 'ns < 0"

thickness. Because the effective London penetration depth * = 2 /d increases considerably in films with thickness d ≪ , the vortex state can √ appear in thin films made from a material with  < 1/ 2. In this case one introduces the effective Ginzburg–Landau parameter ∗ = */, which √ defines the type of superconductivity: type I when ∗ < 1/ 2 √ and type II when ∗ > 1/ 2. In mesoscopic samples the distinction between type I and type II superconductors is determined not only by  and the thickness d, but also by the lateral dimensions of the sample (see, e.g., [16, 34]). In confined mesoscopic samples there is a competition between the triangular Abrikosov distribution of vortices, being the lowest energy configuration in bulk material and thin films, and the sample boundary, which tries to impose its geometry on the vortex distribution. For example, a circular disk will favor vortices situated on a ring near the boundary and only far away from the boundary does its influence diminish and the triangular lattice may reappear.

Multivortex versus Giant Vortex State Depending on the geometry, the size, the applied field, and the temperature, different kinds of vortex states can nucleate in mesoscopic samples: giant vortex states and multivortex states or a mixture of both of them, for example, in triangular and square superconductors (see, e.g., [25]). The multivortex state in mesoscopic confined samples is the analogon of the Abrikosov vortex state in bulk superconductors. The flux penetrates the sample at several positions where vortices are created. These vortices can be very close to each other and overlap, but they are defined by their separate zeros of the Cooper pair density. On the other hand, when the sample is sufficiently small, the vortices will overlap so strongly that it is more favorable to form one giant vortex, corresponding with only one minimum in the Cooper pair density. The shape of the giant vortex also depends on the sample geometry. Figure 6a and b shows the Cooper pair density for a multivortex state and a giant vortex state in a superconducting disk with radius R/ = 6"0. High Cooper pair density is given by dark regions and low by light regions. This means that in Figure 6a the white spots are the vortices.

Vorticity 3.2. Mesoscopic Superconductors In thin superconducting films the distinction between type I and type II superconductivity depends not only on the Ginzburg–Landau parameter  but also on the sample –M

–M

Meissner State

Meissner State

Hc (a) Type-I

H0

Mixed State

Hc2

Hc1

H0

(b) Type-II

Figure 5. The magnetization as a function of the applied magnetic field for type I and type II bulk superconductors.

For a given mesoscopic sample, different superconducting states (giant vortex states and multivortex states) can nucleate for a particular magnetic field. These states have a different free energy and a different vortex configuration, and they can be characterized by their vorticity L. For multivortex states the vorticity is nothing other than the number of vortices. To determine the vorticity of the giant vortex state one has to look at the phase of the order parameter. Along a closed path around the vortex, the phase of the order parameter always changes with L times 2. Figure 6c shows the contour plot of the phase of the order parameter for the multivortex state of Figure 6a. Light regions indicate phases near zero and dark regions phases near 2. By encircling the disk near the boundary of the disk, the phase changes 5 times with 2. This means that the total vorticity of the disk is L = 5. By going around one single vortex the phase changes with 2 and L = 1. In Figure 6d the phase of the

592

Doubly Connected Mesoscopic Superconductors

(a)

5. GIANT VORTEX STATES IN RINGS

(b)

(c)

In small superconducting rings, the sample boundary imposes its symmetry on the vortex configuration. Therefore, the vortex configuration is axially symmetric and only giant vortex states nucleate. In a first step, Bruyndoncx et al. [38] studied the superconducting/normal transition of superconducting rings with zero thickness. To this aim, they were allowed to simplify the problem by linearizing the first Ginzburg–Landau equation and neglecting the second one. Later, we solved the complete set of nonlinear Ginzburg– Landau equations assuming axial symmetry [39].

(d)

5.1. Linear Theory

Figure 6. The Cooper pair density for the multivortex state (a) and the giant vortex state (b) and the phase of the order parameter for the multivortex state (c) and the giant vortex state (d) with vorticity L = 5 in a superconducting disk with radius R/ = 6"0. High (low) Cooper pair density is given by red (blue) regions. Phases near 2 (0) are given by dark (light).

Bruyndoncx et al. [38] studied the phase boundary Tc H of superconducting disks with a hole in the center. Concentrating on the phase boundary they used the linear Ginzburg– Landau theory and neglected the second GL equation which describes the demagnetization effects. The linearized first Ginzburg–Landau equation was solved to find Tc H :   1 2e  2  −i  − A  = − 2m∗ c

(23)

The eigenenergies can be written as order parameter is shown for the giant vortex configuration of Figure 6b. By going around the giant vortex, the phase of the order parameter changes 5 times with 2, which means that the giant vortex state has vorticity L = 5.

4. LITTLE–PARKS EFFECT In 1962 Little and Parks [35, 36] measured the shift of the critical temperature Tc H of a thin-walled Sn microcylinder in an axial magnetic field H . The Tc H phase boundary showed a periodic component, with the magnetic period corresponding to the penetration of a superconducting flux quantum -0 = hc/2e. The Little–Parks oscillations in Tc H are a straightforward consequence of the fluxoid quantization constraint, which was predicted by London [37]. Inserting the order parameter  =  exp i into the current operator we obtain   e 2e  j = A  2   − m c

C

mc     j · d l + A · d l = L-0 2e2  2 C

  2 2 T = 1 − 2m∗  2 T 2m∗  2 0 Tc0

(22)



 · d l = rotA  · d S = H  · d S = - using the where C A Stokes theorem. In other words, when the external flux - is not an integer times the flux quantum -0 , induced supercurrents are necessary to fulfill Eq. (22). The integer number L is the vorticity.

(24)

where Tc0 is the critical temperature in the zero magnetic field. The boundary condition for the order parameter is given by Eq. (15), which expresses the fact that the supercurrent does not have a component perpendicular to the sample boundary. For loop geometries the cylindrical coordinate system  = Hr/2e are used, where e is the r)  and the gauge A tangential unit vector. They found that the exact solution of the first Ginzburg–Landau equation has the following form: - L/2 - exp − -0 2-0   × K −n) L + 1) -0

 -)  = e−iL

(21)

which after integration over a closed contour C inside the superconductor leads to

− =

(25)

where K a) c) y = c1 M a) c) y + c2 U a) c) y with M a) c) y and U a) c) y as the two confluent hypergeometric functions and - = Hr 2 is the applied magnetic flux through a circle of radius r. The integer n determines the energy eigenvalue. The sample topology determines the values of c1 , c2 , and n through the boundary condition. The eigenenergies of the first Ginzburg–Landau Eq. (23) are − =

  1 2eH 2n + 1 = 4 n + 2m∗ 2

(26)

593

Doubly Connected Mesoscopic Superconductors

-0

(27)

where - = Hro2 . The bulk Landau levels can be found when n = 0) 1) 2) " " " is substituted in Eqs. (26) and (27), meaning that the lowest level n = 0 corresponds to the upper critical field Hc2 T = -0 /72 2 T 8. Note that the lowest Landau level n = 0 for a bulk superconductor is degenerate in the phase winding number L, and therefore the eigenfunction can be expanded as  = cL L . Interference patterns between the different functions L give rise to aperiodic vortex states [40]. The boundary condition can be written in cylindrical coordinates as  9 r   =0 (28) 9r r=r0 ) ri

with ri and ro , respectively, as the inner and outer radius of the ring. Using dM a) c) y /dy = a/cM a + 1) c + 1) y and dU a) c) y /dy = −aU a + 1) c + 1) y for the derivates of the first and second types of Kummer functions and inserting Eq. (25) into Eq. (28) gives     c1 L − M −n) L + 1) -0 -0   2n M −n + 1) L + 2) − L + 1 -0 -0     U −n) L + 1) + c2 L − -0 -0     = 0 (29) + 2n U −n + 1) L + 2) -0 -0 r=ro ) ri

which has to be solved numerically for each integer value of L, resulting in a set of values n L) - with - = Hr02 . Notice that one can choose c1 = 1. Figure 7 shows the Landau level scheme (dashed curves) calculated from Eqs. (27) and (29) for loops with two different inner radii: x = 0"5 (a), and x = 0"9 (b). The applied magnetic flux - = ;0 Hro2 is defined with respect to the outer sample area. The Tc H boundary is composed of  solutions with a different phase winding number L and is drawn as a solid cusp-like curve in Figure 7. At - ≈ 0, the state with L = 0 is formed at Tc H and one by one consecutive flux quanta L enter the loop as the magnetic field increases. Each state L approximately has a parabolic dependence, close to Tc H . Like for the disk [1, 4, 41], where - ≈ -0 L + L1/2 , one has L-0 ≤ - here as well, indicating the overall diamagnetic response of the sample. As x increases, the oscillations in Tc - change from cusplike to very pronounced local extrema for x = 0"9. In the limit of vanishing wire width (x → 1), L is the integer closest to -m /-0 , and therefore the response of the loop is alternating between diamagnetic and paramagnetic, as the flux

27

H

24 21 18

ro2/ξ2(T)

∗ = 5 Hc3

(a)

x=ri /ro=0.5

15 12 9 6 3 0 0

2

4

6

8

10

12 14

16

18

20

Φ/Φ0 (b) 2.0 1.8 1.6 1.4

ro2/ξ2(T)

where 4 = 2eH /m∗ is the cyclotron frequency. The parameter n depends on the vorticity L. With Eq. (24) this can be rewritten as     r02 1 r02 Tc H = 4 n + = 1 −  2 Tc  2 0 Tc0 2 -0

1.2 1.0 0.8 0.6 0.4

H

0.2

x=ri /ro=0.9

0.0 0

2

4

6

8

10

12

14

16

18

20

Φ/Φ0

Figure 7. Calculated energy level scheme (dashed curves) for a superconducting loop with different ratio of inner to outer radius x = ri /ro : (a) x = 0"5 and (b) x = 0"9. The lowest level for each magnetic flux -/-0 corresponds to the highest possible temperature Tc H for which superconductivity can exist. A state with phase winding number L = 0 is formed at Tc - ≈ 0 , and at each cusp in Tc H the system makes a transition L → L + 1, indicating the entrance of an extra vortex. The solid and dotted lines correspond to Hc2 T and Hc3 T , respectively. Reprinted with permission from [38], V. Bruyndoncx et al., Phys. Rev. B 60, 10468 (1999). © 1999, American Physical Society.

varies. The solid and dotted straight curves in Figure 7 are the bulk upper critical field Hc2 T and the surface critical field Hc3 T for a semi-infinite slab, respectively. The energy levels below the Hc2 line (solid straight curve in Fig. 7) could be found by fixing a certain L and solving Eq. (29) for a small - until a set n) c2 is found with n < 0. These values were always put in as starting values for a slightly higher -. A trivial solution of Eq. (29) is obtained for n = 0) 1) 2) " " " . Both confluent hypergeometric functions reduce to M −N ) L + 1) -/-0 = 1 and U −N ) L + 1) -/-0 = 1, and thus c2 = −c1 . Inserting this into Eq. (25) gives  -)  = 0 everywhere. With this method, one is able to find solutions with n < 0 numerically. Note that the lowest Landau level always has a lower energy − -  than that for a semi-infinite superconducting slab. The dash–dotted curve in Figure 7 gives the result obtained with the London limit [42]. For x = 0"5 (Fig. 7a) the London limit is valid up to - ≈ 8-0 and for x = 0"9 (Fig. 7b) the assumption of a spatially constant   gives a Tc - , which cannot be distinguished from the exact solution of Eq. (23).

594

Doubly Connected Mesoscopic Superconductors

Next, the periodicity >- of Tc - is discussed for the different L states in the loops. These results are shown in Figure 9. For the disk the first transition occurs at - = 1"92-0 , so >- L = 0 = 3"85-0 . The period >- goes down for increasing L, until it reaches the asymptotic limit [2]: >- = -0 1 + 2 ?-/-0 −1/2

(30)

The filled squares are the periods for the filled disk. The periodicity of Tc - in the case of loops behaves differently: for x = 0"1 (filled up-triangles in Fig. 9) >- L = 1 = 1"82-0 is larger than for the disk, and then >- L = 2 jumps below the corresponding value for the disk; and for higher L ≥ 3, the period >- matches the disk behavior. Consequently, the giant vortex state builds up for L ≥ 3; that is, -  5-0 . A similar analysis can be carried out for a loop with x = 0"5 (filled down-triangles in Fig. 9). For low L, >- ≈ 1"8-0 , and then >- L decreases substantially below the value for the disk, before increasing again until the same period >- L is reached for L ≥ 12. The loop is in the giant vortex state when -  17-0 . For loops made of even thinner wires (x = 0"7 and x = 0"9) (open symbols in Fig. 9), Tc - stays periodic up to L = 13. The constant period >corresponds to >-m = 1 + x /2 2 >- = 1-0 , as it should be in the London limit for x → 1.

5.2. Nonlinear Theory But what happens away from the superconducting/normal transition, that is, deeper inside the superconducting state? To study this we solved the complete set of Ginzburg– Landau equations, assuming axial symmetry, which implies that the Ginzburg–Landau equations are solved for a fixed value of the angular momentum or vorticity L. Therefore, the order parameter has the form  @) # = f @ eiL#

(31)

1.9 Disk × = 0.1 × = 0.5 × = 0.7 × = 0.9

1.8 1.7 1.6 1.5

∆Φ/Φ0

For x = 0"5 (Fig. 7a) Tc H has a rather parabolic background for flux -  14-0 and becomes quasi-linear at higher flux. In Figure 7b the background is parabolic in the entire flux regime. Simultaneously, as x increases, the cusps in Tc H become more and more pronounced, until the usual Little–Parks effect is recovered for x = 0"9, where sharp local minima and clear maxima in Tc H are seen. The   profiles for a loop with x = 0"5 are shown in Figure 8 for different values of vorticity. # is chosen to be equal to 9-0 , and the normalization  ro  = 1. For L = 2 (Fig. 8a),   has a maximum at r = ri . For higher L the order parameter distribution flattens until it reaches L = 4. Then, for the ground state energy (L = 5) (Fig. 8b), the maximum in   moves outward, but we should notice that superconductivity nucleates in a quasi-uniform way. Indeed, the exact Tc - and the London limit result are still very close to each other at - = 9-0 (see Fig. 7a). Although multiple flux quanta L are threading the middle opening of the loop, we cannot, in the strict sense, speak about a giant vortex state here. First of all, there is no real “normal core” within the sample area, and secondly we are not dealing with a surface superconducting state in this case. The surface-to-volume ratio is so large that the whole sample area becomes superconducting at once. On the contrary, for the disk, strong spatial gradients of   are responsible for the spontaneous breaking of superconductivity in the giant vortex core, while only a surface sheath is superconducting. It is worth noting that for all of the states shown in Figure 8 with x = 0"5 the different L states correspond to w < 2"1 T (see also Fig. 7a). At this point we want to remind the reader that in a thin film of thickness d in a parallel field H , a dimensional crossover is found at d = 1"84 T . For low fields (high ) Tc H is parabolic [twodimensional (2D)], and for higher fields vortices start penetrating the film and consequently Tc H becomes linear [three-dimensional (3D)] [43]. In Figure 7a the arrow indicates the point on the phase diagram Tc - where w = 1"84 T . For the loop with x = 0"9 this point lies outside the flux regime of the calculation. A dimensional transition from 2D to 3D shows up approximately at this point, although the vortices are not penetrating the sample area in the 3D regime. Instead, the middle loop opening contains an integer number of flux quanta L-0 . In the present case, the 2D–3D crossover roughly occurs at the value - where the London limit result fails to describe the exact Tc - .

1.4 1.3 1.2 1.1 1.0

2.5 1.75 1

1 0.8 0.6

1.05 1 0.95

0.9 0.8 0.7 0

1

2

3

4

5

6

7

8

9 10 11 12 13

L

a) L = 2

b) L = 5

c) L = 6

Figure 8. Order parameter distribution   for x = 0"5, at a fixed - = 9 -0 . (a) L = 2, (b) L = 5, and (c) L = 6. The state with L = 5 is only slightly modulated (  ≈ constant: A = 1"01) and corresponds to the ground state level  − LLL 9-0 . Reprinted with permission from [38], V. Bruyndoncx et al., Phys. Rev. B 60, 10468 (1999). © 1999, American Physical Society.

Figure 9. The period >- of the phase boundary Tc - in units of the flux quantum -0 as a function of the phase winding number L. The data for several of the loops are shown as a symbol and are compared to the period in a filled disk (filled squares). The interconnecting lines are only a guide to the eye. Reprinted with permission from [38], V. Bruyndoncx et al., Phys. Rev. B 60, 10468 (1999). © 1999, American Physical Society.

595

Doubly Connected Mesoscopic Superconductors

and −2



9 1 9@A 9 2 A + 9@ @ 9@ 9z2



=



     L @ 2z − A f 2C C @ R d (33)

 = Ae# , where C x = 1 for x < 1 and C x = 0 for x > 1, A and  means averaging over the disk thickness g r  = 1 +d/2 g z) r dz. d −d/2 The boundary condition for the order parameter (15) can be written as   9f  9f  = 0) =0 (34) 9@ @=Ro 9@ @=Ri

where Ri and Ro are the inner and outer radius of the ring. To solve the system of Eqs. (32) and (33) numerically, a finite difference representation is applied on the space grid @i ) zi . Because the size of the simulation region exceeds by far the size of the disk, this space grid will be nonuniform outside the disk region to diminish computation time. Outside the disk region, the grid space increases exponentially with the distance. Because of this, it becomes possible to use the same number of grid points inside and outside the disk. Integrating over the line @i−1/2 < @ < @i+1/2 , where @i+1/2 = @i+1 + @i /2, the first Ginzburg–Landau equation (32) becomes   fi+1 − fi fi − fi−1 2 @ − @ − 2 i+1/2 i−1/2 @i+1 − @i @i − @i−1 @i+1/2 − @2i−1/2  2  L fi = fi − fi3 (35) −A + @ i where fi = f @i . The steady-state solution of the Ginzburg– Landau equations was obtained using an iteration procedure. Therefore, an upper index k is introduced to denote the iteration step. To speed up the convergency an iteration parameter ?f is added and the nonlinear term is expanded in fik 3 = fik−1 3 + 3 fik−1 2 fik − fik−1 , which leads to the first Ginzburg–Landau equation in finite difference form:   k k fi+1 − fik fik − fi−1 2 ?f fik − 2 − @ @ i−1/2 i+1/2 @i+1 − @i @i − @i−1 @i+1/2 − @2i−1/2  2  L fik − fik + 3 fik−1 2 fik −A + @ i = ?f fik−1 + 2 fik−1 3

(36)

To obtain the finite difference representation of the second Ginzburg–Landau Eq. (32), we integrated this equation over the square zj−1/2 < z < zj+1/2 , @i−1/2 < @ < @i+1/2 . By introducing the upper index k denoting the iteration step and the iteration parameter ?a to speed up the convergence,

the finite difference representation of the second Ginzburg– Landau equation becomes   @i Akj)i − @i−1 Akj)i−1 @i+1 Akj)i+1 − @i Akj)i 22 − 2 2 2 2 @i+1/2 − @i−1/2 @i+1 − @i @i − @i−1  k  k k k 2 A A − A − A 2 j)i j−1)i j+1)i j)i − − zj+1/2 − zj−1/2 zj+1 − zj zj − zj−1   L − Akj)i fik 2 = ?a Ak−1 − i)j @i

?a Akj)i −

(37)

The iteration parameters ?f and ?a correspond to an artificial time relaxation of the system to a steady state with time steps 1/?f and 1/?a . For typical values of the iteration parameters ?f = 2, ?a = 5 convergency is reached after a few hundred iteration steps. The ground state free energy F of a superconducting disk with radius Ro = 2"0 and thickness d = 0"005 and GL parameter  = 0"28 is shown in Figure 10 for a hole in the center with radius Ri / = 0"0, 0"5, 1"0, and 1"5. The situation with Ri = 0 corresponds to a superconducting disk without a hole, which was already studied in [13]. With increasing hole radius Ri , we find that the superconducting/normal transition (this is the magnetic field where the free energy equals zero) shifts appreciably to higher magnetic fields, and more transitions between different L states are possible before superconductivity disappears. Furthermore, the thin dotted curve gives the free energy for a thicker ring with thickness d = 0"1 and Ri = 1"0. In comparison with the previous results for d = 0"005, the free energy becomes more negative, but the transitions between the different L states occur almost at the same magnetic fields. Thus increasing the thickness of the ring increases superconductivity, which is a consequence of the smaller penetration of the magnetic field into the superconductor. Experimentally, using magnetization measurements one can investigate the effect of the geometry and the size of the sample on the superconducting state. To investigate the magnetization of a single superconducting disk Geim et al. 0.0

d/ξ = 0.1 –0.2

F/F0

and, consequently, both the vector potential and the superconducting current are directed along e# . For a fixed angular momentum L, the Ginzburg–Landau equations can be reduced to  2  L 1 9 9f @ + − A f = f 1 − f 2 (32) − @ 9@ 9@ @

Ri/ξ = 1.0

Ro /ξ = 2.0 d/ξ = 0.005 κ = 0.28

–0.4 –0.6

Ri /ξ = 0.0 Ri /ξ = 0.5 Ri /ξ = 1.0

–0.8

Ri /ξ = 1.5

–1.0 0

1

2

3

4

5

6

7

8

H0 /Hc2 Figure 10. The ground state free energy as a function of the applied magnetic field H0 of a superconducting disk with radius Ro = 2"0, thickness d = 0"005, and  = 0"28 for a hole in the center with radius Ri / = 0"0, 0"5, 1"0, and 1"5, respectively. The thin dotted curve gives the free energy of a thicker ring with thickness d = 0"1 and Ri = 1"0. The free energy is in units of F0 = Hc2 V /8. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

596

Doubly Connected Mesoscopic Superconductors

[1] used submicron Hall probes. The Hall cross acts like a magnetometer, where in the ballistic regime the Hall voltage is determined by the average magnetic field through the Hall cross region [44]. Hence, by measuring the Hall resistance, one obtains the average magnetic field and, consequently, the magnetic field expelled from the Hall cross [Eq. (20)], which is a measure for the magnetization of the superconductor [14]. As for a superconducting disk [14], the field distribution for thin superconducting rings is extremely nonuniform inside as well as outside the sample and therefore the detector size will have an effect on the measured magnetization. To understand this effect of the detector, we calculate the magnetization for a superconducting ring with outer radius Ro = 2"0, thickness d = 0"005, and two values of the inner radius Ri = 0"5 and Ri = 1"5 by averaging the magnetic field [see Eq. (20)] over several detector sizes S. The results are shown in Figure 11a and b as a function of the applied magnetic field for Ri = 0"5 and Ri = 1"5, respectively. The solid curve [curve (i)] shows the calculated magnetization if we average the magnetic field over the superconducting volume, and curve (ii) is the magnetization after averaging the field over a circular area with radius Ro , that is, superconductor and hole. Notice that in the Meissner regime, that is, H0 /Hc2 < 0"7 where L = 0, the magnetization of superconductor + hole is larger

–M/Hc2 (10–4)

(a)

6 5

(ii)

4

(i)

Ro /ξ = 2.0 d/ξ = 0.005 κ = 0.28

3

(iii)

2 1

(iv)

0

Ri /ξ = 0.5

–1 0.0

0.5

1.0

1.5

2.0

2.5

than that of the superconductor alone, which is due to flux expulsion from the hole. For L ≥ 1 the reverse is true because now flux is trapped in the hole. Experimentally, one usually averages the magnetic field over a square Hall cross region. Therefore, we calculated the magnetization by averaging the magnetic field over a square region [curve (iii)] with width equal to 2Ro , that is, the diameter of the ring. Curve (iv) shows the magnetization if the sides of the square detector are equal to 2 + 1/2 Ro . Increasing the size of the detector decreases the observed magnetization because the magnetic field is averaged over a larger region, which brings H  closer to the applied field H0 . Having the free energies of the different giant vortex configurations for several values of the hole radius varying from Ri = 0"0 to Ri = 1"8, we construct an equilibrium vortex phase diagram. Figure 12 shows this phase diagram for a superconducting disk with radius Ro = 2"0, thickness d = 0"005, and  = 0"28. The solid curves indicate where the ground state of the free energy changes from one L state to another, and the thick solid curve gives the superconducting/normal transition. The latter exhibits a small oscillatory behavior, which is a consequence of the Little–Parks effect. Notice that the superconducting/normal transition is moving to higher fields with increasing hole radius Ri and more flux can be trapped. In the limit Ri → Ro , the critical magnetic field is infinite, and there are an infinite number of L states possible, which is a consequence of the enhancement of surface superconductivity for very small samples [15, 18, 45–47]. Because of the finite grid, we were not able to obtain accurate results for Ri ≈ Ro . The dashed lines connect our results for hole radius Ri = 1"8 with the results for Ri → Ro [48], where the transitions between the different L states occur when the enclosed flux is # = L + 1/2 #0 , where #0 = ch/2e is the elementary flux quantum. Notice that for rings of nonzero width, that is, Ri = Ro ) the L → L + 1 transition occurs at higher magnetic field than predicted from the condition # = L + 1/2 #0 . The discrepancy

H0 /Hc2 (b)

φ/φ0

5

2.0

0 1

2

4 3

4

6 5

8 6

7

10 8

12

14

9 10

16 15

3 1.5

2 1

Ri /ξ

–M/Hc2 (10–4)

4

2

0

Superconducting state

Normal state

1.0

–1

Ri /ξ = 1.5

–2 0

2

4

6

0.5

Ro /ξ = 2.0 d/ξ = 0.005 κ = 0.28

8

H0 /Hc2

L= 0 0.0 0

Figure 11. The magnetization as a function of the applied magnetic field H0 for a superconducting disk with radius Ro = 2"0 and thickness d = 0"005 and  = 0"28 for a hole in the center with radius Ri / = 0"5 (a) and Ri / = 1"5 (b). Curve (i) is the calculated magnetization if we average the magnetic field over the superconducting volume, curve (ii) after averaging the field over the area with radius Ro , that is, superconductor and hole, and curves (iii) and (iv) after averaging the magnetic field over a square region with widths equal to 2Ro and 2 + 1/2 Ro , respectively. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

1

2

3

4

5

6

7

8

H0/Hc2

Figure 12. Phase diagram: the relation between the hole radius Ri and the magnetic fields H0 at which giant vortex transitions L → L + 1 takes place for a superconducting disk with radius Ro = 2"0 and thickness d = 0"005 and for  = 0"28. The solid curves indicate where the ground state changes from one L-state to another one and the thick solid curve gives the superconducting/normal transition. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

597

Doubly Connected Mesoscopic Superconductors

2.0

d/ξ

1.5

Superconducting

Normal

State

State

1.0

0.5

Ro/ξ = 2.0 Ri /ξ = 0.5 κ = 0.28 L=0

1

2

0.0 0

1

2

3

4

H0 /Hc2

Figure 13. The same as Figure 12 but for varying thickness d of the ring for fixed Ro = 2"0, Ri = 0"5, and  = 0"28. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

0.0 6,7 5

–0.2 2

Ro/ξ = 2.0

F/F0

–0.4

Ri /ξ = 0.5 –0.6

1,3

d/ξ = 0.1 κ = 0.28

4

–0.8

(a)

–1.0 12 1 2

5

8

–M/Hc2 (10–3)

increases with increasing width of the ring and with increasing L. Starting from Ri = 0 we find that with increasing Ri the Meissner state disappears at smaller H0 . The hole in the center of the disk allows for a larger penetration of the magnetic field which favors the L = 1 state. This is the reason why the L = 0 → L = 1 transition moves to a lower external field while the L = 1 → L = 2 transition initially occurs for larger H0 with increasing Ri . When the hole size becomes of the order of the width of one vortex the L = 1 → L = 2 transition starts to move to lower fields and the L = 2 state becomes more favorable. The effect of the thickness of the ring on the phase diagram is investigated in Figure 13. The thin solid curves indicate where the ground state of the free energy changes from one L state to another, while the thick curve gives the superconducting/normal transition. Notice that the transition from the L = 0 to the L = 1 state and the superconducting/normal transition depends weakly on the thickness of the ring. For increasing thickness d) the L = 2 state becomes less favorable and disappears for d  0"7. In this case, there is a transition from the L = 1 state directly to the normal state. Increasing d stabilizes the different L states up to larger magnetic fields. This is due to the increased expulsion of the applied field from the superconducting ring [15]. In the next step, we investigate three very important and mutually dependent quantities: the local magnetic field H , the Cooper pair density  2 , and the current density j. We will discuss these quantities as a function of the radial position @. For this study we distinguish two situations; that is, Ri ≪ Ro and Ri  Ro . In the first case the sample behaves more like a superconducting disk and in the second case like a superconducting loop. First, we consider a superconducting disk with radius Ro = 2"0 and thickness d = 0"1 with a hole with radius Ri = 0"5 in the center. The free energy and the magnetization (after averaging over the superconductor + hole) for such a ring are shown in Figure 14. The dashed curves

6 4 4

7

0 3 –4

(b)

–8 0.0

0.5

1.0

1.5

2.0

2.5

H0 /Hc2

Figure 14. The free energy (a) and the magnetization after averaging over the superconductor + hole (b) as a function of the applied magnetic field for a superconducting disk with outer radius Ro = 2"0 and thickness d = 0"1 with a hole with radius Ri = 0"5 in the center  = 0"28 for the different L states (dashed curves) and for the ground state (solid curves). The open circles are at (1) L = 0, H0 /Hc2 = 0"745; (2) L = 0, H0 /Hc2 = 0"995; (3) L = 1, H0 /Hc2 = 0"745; (4) L = 1, H0 /Hc2 = 0"995; (5) L = 1, H0 /Hc2 = 1"7825; (6) L = 1, H0 /Hc2 = 2"0325; and (7) L = 2, H0 /Hc2 = 2"0325. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

give the free energy and the magnetization for the different L states, and the solid curve is the result for the ground state. Figure 15a and b shows the local magnetic field H, Figure 15c and d the Cooper pair density  2 , and Figure 15e and f the current density j as a function of the radial position @ for such a ring at the L states and magnetic fields as indicated by the open circles in Figure 14a and b. For low magnetic field, the system is in the L = 0 state, that is, the Meissner state, and the flux trapped in the hole is considerably suppressed. Hence, the local magnetic field inside the hole is lower than the external applied magnetic field as is shown in Figure 15a by curves 1 and 2. Please notice that the plotted magnetic field is scaled by the applied field H0 . In the L = 0 state the superconductor expels the magnetic field by inducing a supercurrent that tries to compensate the applied magnetic field in the superconductor and inside the hole. This is called the diamagnetic Meissner effect. As long as L equals zero, the induced current has only to compensate the magnetic field at the outside of the ring and, therefore, the current flows in the whole superconducting material in the same direction and the size of the current increase with increasing field. This is clearly shown in Figure 15e by curves 1 and 2, where the current density j becomes more negative for increasing H0 /Hc2 . Notice also that the current density is more negative at the outside than

598

Doubly Connected Mesoscopic Superconductors

(a)

2.0

(b)

1.2

1.8 1.1

1.4

H/H0

H/H0

1.6 3 1.2 4

1.0

7

0.9

6

1.0 0.8 2

5

0.6 1 0

2

1

0.8

4

3

1

0

3

2

ρ/ξ

4

ρ/ξ

0.9 0.6

1

0.5

0.8

|Ψ|2

|Ψ|2

2

0.7 4

1.2

0.3

(c)

0.1

(e)

0.8 0.6

0.8

6

(d)

7

(f)

5

0.4 6 0.2

0.4

j/j0

j/j0

0.4

0.2

0.6 3

3 4

0.0 –0.4

7

0.0 –0.2

1

0.5

5

–0.4

2

1.0

–0.6 1.5

ρ/ξ

2.0

0.5

1.0

1.5

2.0

ρ/ξ

Figure 15. The local magnetic field H (a and b), the Cooper pair density  2 (c and d) and the current density j (e and f) for the situations indicated by the open circles in Figure 14 as a function of the radial position @ for a superconducting disk with radius Ro = 2"0 and thickness d = 0"1 with a hole with radius Ri = 0"5 in the center  = 0"28 . Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

at the inside of the superconducting ring, which leads to a stronger depression of the Cooper pair density at the outer edge compared to the inner edge of the ring (see curves 1 and 2 in Fig. 15c). At H0 /Hc2 = 0"745, the ground state changes from the L = 0 to the L = 1 state (see Fig. 14a). Suddenly more flux becomes trapped in the hole (compare curve 1 with curve 3 in Fig. 15a), the local magnetic field inside the hole increases and becomes larger than the external magnetic field H0 . In the L = 1 state, there is a sharp peak in the magnetic field at the inner boundary because of demagnetization effects. Consequently, more current is needed to compensate the magnetic field near the inner boundary than that near the outer boundary (see curve 3 in Fig. 15e). The sign of the current near the inner boundary becomes positive (the current direction reverses), but the sign of current near the outer boundary does not change. This can be explained as follows. Near the inner boundary @  1"0 the magnetic field is compressed into the hole (paramagnetic effect), whereas near the outer boundary @  2"0 the magnetic field is expelled to the insulating environment (diamagnetic effect). The sign reversal of j occurs at @ = @∗ , and later we will show that the flux through the circular area with radius @∗ is exactly quantized. At the L = 0 to the L = 1 transition the maximum in the Cooper pair density (compare curves 1 and 3 in Fig. 15c) shifts from @ = Ri to @ = Ro . Further

increasing the external field increases the Cooper pair density near the inner boundary initially (compare curves 3 and 4 in Fig. 15c), because the flux in the hole has to be compressed less. The point @∗ , where j = 0, shifts toward the inner boundary of the ring. By further increasing the external magnetic field, the Cooper pair density starts to decrease (see curves 5 and 6 in Fig. 15d) and attains its maximum near the inner boundary. The current near the inner boundary becomes less positive (see curves 5 and 6 in Fig. 15f), that is, less shielding of the external magnetic field inside the hole (see curves 5 and 6 in Fig. 15b), and near the outer boundary j becomes less negative, which shields the magnetic field from the superconductor + hole. Thus at the outer edge the local magnetic field has a local maximum, which decreases with applied magnetic field H0 . At H0 /Hc2 ≈ 2"0325, the ground state changes from the L = 1 state to the L = 2 state and extra flux is trapped in the hole. The changes in the magnetic field distribution, the Cooper pair density and the current density are analogous to the changes at the first transition. For example, the magnetic field inside the hole increases compared to the external magnetic field (curve 7 in Fig. 15b), the radius @∗ increases substantially (curve 7 in Fig. 15f), and the maximum in the Cooper pair density shifts to the outer boundary (curve 7 in Fig. 15d). For Ri ≪ Ro and L > 0, the superconducting state consists of a combination of the paramagnetic and the diamagnetic Meissner state, like for a disk. For Ri  Ro we expect that the sample behaves like a loop and, hence, the superconducting state is in a pure paramagnetic Meissner state or a pure diamagnetic Meissner state. We consider a superconducting disk with radius Ro = 2"0 and thickness d = 0"1 with a hole with radius Ri = 1"8 in the center. The free energy and the magnetization (after averaging over the superconductor + hole) for such a ring are shown in Figure 16. The dashed curves give the free energy and the magnetization for the different L states, and the solid curve is the result for the ground state. Figure 17a and b shows the local magnetic field H and Figure 17c and d the current density j as a function of the radial position @ for such a ring at the L states and magnetic fields as indicated by the open circles in Figure 16a and b. The Cooper pair density has almost no structure and is practically constant over the ring and will, therefore, not be shown. For L = 0, the situation is the same as that for Ri ≪ Ro . The magnetic field is expelled from the superconductor and the hole to the outside of the system, that is, a diamagnetic Meissner effect. The current flows in the whole superconducting material in the same direction (curves 1 and 2 in Fig. 17a) and the size increases with increasing external field H0 (curves 1 and 2 in Fig. 17c). At H0 /Hc2 = 0"27, the ground state changes from the L = 0 state to the L = 1 state and suddenly more flux is trapped in the hole. The local magnetic field inside the hole becomes larger than the external field H0 , and there is a sharp peak near the inner boundary (curves 3 and 4 in Fig. 17a). In contrast to the situation for Ri ≪ Ro , there is no peak near the outer boundary, which means that the magnetic field is only expelled to the hole, that is, a paramagnetic Meissner effect. The induced current flows in the reverse direction in the whole superconductor (curves 3 and 4 in Fig. 17c). For increasing external magnetic field, the

599

Doubly Connected Mesoscopic Superconductors 1.2

1.6

(a) 1.4

Ri /ξ = 1.8

1.2

1.1

F/F0

d/ξ = 0.1 –0.4

(b)

κ = 0.28

H/H0

–0.2

Ro /ξ = 2.0 H/H0

0.0

3

1.0

4

0.8

1

2

7

1.0 5 6

0.9

–0.6

0.6

0

2

1

2

4

3

0

1

ρ/ξ

–0.8

2

3

4

ρ/ξ

6,7

1,3

4

5

–1.0

0.2

(a)

0.1

(c)

3

0.3

(d)

7

0.2

4

0.1 2 6

1

–0.1 –0.2

2

–M/Hc2 (10–3)

0.0

–0.3

5

1.8

1

7

–4

(b) 0.5

1.0

1.5

5

–0.2 2

1.9

2.0

–0.3 1.8

6

1.9

2.0

ρ/ξ

Figure 17. The local magnetic field H (a and b) and the current density j (c and d) for the situations indicated by the open circles in Figure 16 as a function of the radial position @ for a superconducting disk with radius Ro = 2"0 and thickness d = 0"1 with a hole with radius Ri = 1"8 in the center  = 0"28 . Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

4

–2 3

0.0 –0.1

ρ/ξ

0

0.0

j/j0

j/j0

4

2.0

H0 /Hc2

Figure 16. The free energy (a) and the magnetization after averaging over the superconductor + hole (b) as a function of the applied magnetic field for a superconducting disk with outer radius Ro = 2"0 and thickness d = 0"1 with a hole with radius Ri = 1"8 in the center  = 0"28 for the different L states (dashed curves) and for the ground state (solid curves). The open circles are at (1) L = 0, H0 /Hc2 = 0"27; (2) L = 0, H0 /Hc2 = 0"395; (3) L = 1, H0 /Hc2 = 0"27; (4) L = 1, H0 /Hc2 = 0"395; (5) L = 1, H0 /Hc2 = 0"695; (6) L = 1, H0 /Hc2 = 0"82; and (7) L = 2, H0 /Hc2 = 0"82. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

magnetic field inside the hole, the height of the demagnetization peak, and hence the size of the current decrease (see curves 3 and 4 in Fig. 17a and c). With a further increase in the field, the superconducting state transforms into a diamagnetic Meissner state. The magnetic field is now expelled to the outside of the sample (curves 5 and 6 in Fig. 17b) and the direction of the current is the same everywhere in the ring (curves 5 and 6 in Fig. 17d). At H0 /Hc2 = 0"82, the ground state changes from the L = 1 state to the L = 2 state. The changes in the magnetic field distribution (see curve 7 in Fig. 17b) and the current density (see curve 7 in Fig. 17c) are analogous to the changes at the first transition. The diamagnetic state transforms into a paramagnetic state.

6. FLUX QUANTIZATION Zharkov and Arutunyan [49, 50] found that in a thin-walled hollow cylinder the flux is not quantized through the hole, but through a circular area with (effective) radius equal to the geometric mean square of the inner and outer radius of

 the hollow cylinder; that is, @∗ = Ri Ro . These results were obtained within the London limit. We investigated the flux quantization in the fat ring. Figure 18 shows the flux through the hole (a) and through the superconductor + hole (b) as a function of the applied magnetic field H0 for a superconducting disk with radius Ro = 2"0 and thickness d = 0"1 with a hole in the center with radius Ri = 1"0. The dashed curves show the flux for the different L states, the solid curve for the ground state, and the thin solid line is the flux through the hole if the sample is in the normal state. It is apparent that the flux through the hole (or through the superconducting ring + hole) is not quantized (see Fig. 18a). At H0 /Hc2 = 0"4575 suddenly more flux enters the hole and the ground state changes from the L = 0 state to the L = 1 state. At this transition also the flux increase through the hole is not equal to one flux quantum #0 . It is generally believed that the flux through a superconducting ring is quantized. But as was shown in [42], this is even no longer true for hollow cylinders when the penetration length is larger than the thickness of the cylinder wall. The present result is a generalization of this observation to mesoscopic ring structures. Note that for the case in Figure 18 the penetration length is / = 0"28 and the effective penetration length */ = 0"78 is comparable to the width of the ring Ro − Ri / = 1"0. For L > 0, the superconducting current equals zero at a certain “effective” radial position @ = @∗ . It is the flux through the circular area with radius @∗ that is quantized and not necessarily the flux through the hole of our disk. Inserting the order parameter  =   exp i into the current operator, we obtain   2e  e A (38)  2   − j = m c

600

Doubly Connected Mesoscopic Superconductors 1.0

2.0

Ro /ξ = 2.0

0.8

Ro /ξ = 2.0

Ri /ξ = 1.0

0.6

Ri /ξ = 1.0

d/ξ = 0.1

0.4

d/ξ = 0.1

0.2

κ = 0.28

κ = 0.28 1.0

j/j0

φ/φ0

1.5

(a)

0.0 –0.2

0.5 –0.4

(a)

–0.6

H0 /Hc2 = 1.6075

0.0 3 L=1

3

L=2 L=3

φ/φ0

φ/φ0

2 2 –3.5

1 1

(b) 0 1

0

2

3

H0/Hc2

Figure 18. The flux through (a) the hole and (b) the superconductor + the hole as a function of the applied magnetic field H0 for a superconducting disk with radius Ro = 2"0 and thickness d = 0"1 with a hole in the center with radius Ri = 1"0. The dashed curves show the flux for the different L states and the solid curve for the ground state. The thin solid line is the flux in the absence of superconductivity. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

which after integration over a closed contour C inside the superconductor leads to   mc  · d l = L#0 + A j (39) 2 2 C 2e   When the contour C is chosen along a path such that the superconducting current is zero we obtain, using Stokes’ theorem,    · d l = rotA  · d S = H  · d S = # (40) L#0 = A C

which tells us that the flux through the area encircled by C is quantized. In our wide superconducting ring the current is nonzero at the inner boundary of the ring and, consequently, the flux through the hole does not have to be quantized. To demonstrate that this is indeed true, we show in Figure 19a the current density as a function of the radial position @ and in Figure 19b the flux through a circular area with radius @ for a superconducting disk with radius Ro = 2"0 and thickness d = 0"1 with a hole in the center with radius Ri = 1"0 in the presence of an external magnetic field H0 /Hc2 = 1"6075 for the case of three different giant vortex states; that is, L = 1) 2, and 3. In the L = 1 state, the current density equals zero at a distance @∗ / ≈ 1"16 from the center and the flux through an area with this radius is

(b)

0 0.0

0.5

1.0

1.5

2.0

ρ/ξ

Figure 19. The current density (a) and the flux (b) as a function of the radial position @ for a superconducting ring with Ro = 2"0, Ri = 1"0, d = 0"1, and  = 0"28 for L = 1 (solid curves), L = 2 (dashed curves), and L = 3 (dotted curves) at an applied magnetic field H0 /Hc2 = 1"6075. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

exactly equal to one flux quantum #0 . For L = 2 and L = 3, the current density j equals zero at @∗ / ≈ 1"56 and 1"91, respectively, and the flux through the area with this radius @∗ is exactly equal to 2#0 and 3#0 , respectively. We find that @∗ depends on the external applied magnetic field and on the value of L, contrary to the results of Arutunyan and Zharkov [49, 50], which for the case in Figure 20a would give @∗ = 1"41. The dependence of @∗ as a function of the applied magnetic field is shown in Figure 20a. The dashed curves give the @∗ of the different L states. For increasing magnetic field and fixed L, the value of @∗ decreases; that is, the “critical” radius moves toward the inner boundary. The solid curve gives @∗ for the ground state. At the L → L + 1 transitions, @∗ jumps over a considerable distance toward the outside of the superconducting ring, and the size of the jumps decreases with increasing L. The dotted lines in the figure give the two boundaries of the superconducting ring: the outer boundary Ro = 2"0 and the inner boundary Ri = 1"0. Note that in Figure 20 there is no @∗ given for the L = 0 state, because only the external magnetic field has to be compensated so that the current has the same sign everywhere inside the ring. In Figure 20b we repeated this calculation for a hole with radius Ri = 1"5, where superconductivity remains to higher magnetic fields and many more L → L + 1 transitions are possible. The result of [49, 50] gives in this case @∗ = 1"73. Note that the results for the L = 1 and the L = 2 states are not connected. The reason is that just before the L = 1 → L = 2 transition the critical

601

Doubly Connected Mesoscopic Superconductors

(a)

2.0

Ro

Ri/ξ = 1.0

1.8

2.5

Ro /ξ = 2.0 Ri /ξ = 1.8

2.0

L=4

φ/φ0

ρ*/ξ

d/ξ = 0.1 1.6 1.4

κ = 0.28

1.0

Ro /ξ = 2.0

L=3

d/ξ = 0.1 κ = 0.28

1.2

1.5

0.5

L=2

Ri

L=1

1.0

(a) 0.0

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.4

3.5

R = Ri

H0 /Hc2 0.2

(b)

Ro

Ri/ξ = 1.5

2.0

j(R)/j0

2.1

0.0 R = Ro

–0.2

1.9

ρ*/ξ

10

1.8

–0.4

(b) 1.7 9 8

1.6

7

–0.90

6 2

4

3

Ri

5

F/F0

1.5 L=1

–0.95

1.4 0

1

2

3

4

5

6

7

H0 /Hc2

Figure 20. The dependence of the effective radius @∗ as function of the applied magnetic field for a superconducting disk with radius Ro = 2"0 and thickness d = 0"1 with a hole in the center with radius Ri = 1"0 (a), and Ri = 1"5 (b). The dashed curves show @∗ for the different L states and the solid curve is for the ground state. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

current for the L = 1 state is strictly positive in the whole ring and hence @∗ is not defined. Notice that the results in Figure 20a and b oscillate around the average value @∗ =  Ri R o . For even narrower rings the current in the ring is mostly diamagnetic or paramagnetic and the region in the magnetic field over which both directions of current occur in the ring becomes very narrow. This is illustrated in Figure 21 for a ring with outer radius Ro = 2"0 and inner radius Ri = 1"8. We plot in Figure 21a the flux through the hole, which becomes very close to the external flux, that is, the flux without any superconductor. In Figure 21b the value of the current at the inner and the outer side of the ring is shown, which illustrates nicely the fact that over large ranges of the magnetic field the current in the ring can flow in one direction. The free energy becomes minimal (Fig. 21c) at a magnetic field where the current in the rings can flow in both directions.

7. MULTIVORTEX STATES Until now, we considered only small rings. In such rings, the confinement effect dominates, and we found that only the giant vortex states are stable. Now we will consider larger wide superconducting rings in which multivortex states can

–1.00

(c) 0.0

0.5

1.0

1.5

H0 /Hc2

Figure 21. (a) The flux through the hole of the ring as a function of the applied magnetic field. The thin dashed line is the applied flux. (b) The current at the inner side (solid curve) and at the outer side (dashed curve) of the ring and (c) the free energy of the ground state as function of the applied magnetic field. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

nucleate for certain magnetic fields. This means that we are no longer allowed to assume axial symmetry for the order parameter. For thin disks d ≪ )  it is allowable to average the GL equations over the disk thickness. Using dimensionless  = 0 for the vector variables and the London gauge divA  we write the system of GL equations in the potential A, following form  2  =  1 −  2 −i2D − A = − >3D A

d  z j2D 2

(41) (42)

where 1 ∗   −  2 A j2D =  ∗ 2D  −  2D 2i

(43)

is the density of superconducting current. The superconducting wavefunction satisfies the following boundary conditions    =0 −i2D − A   Ri r=

(44)

   =0 −i2D − A   Ro r=

(45)

602

Doubly Connected Mesoscopic Superconductors

F = V −1



V

 −A  0 · j2d −  4 8d r 72 A

(46)

where integration is performed over the sample volume V )  0 is the vector potential of the uniform magnetic field, and A we find the ground state. The dimensionless magnetization, which is a direct measure of the expelled magnetic field from the sample, is defined as M=

H  − H0 4

0.0

F/F0

–0.2 –0.4

Ro /ξ = 4.0 Ri /ξ = 0.4

–0.6

d/ξ = 0.005 –0.8

κ = 0.28 (a)

–1.0

4

–M/Hc2 (10–4)

 = 1 H0 @e# far away from the superconductor. Here and A 2 the distance is measured in units of the coherence length , the vector potential √ in c/2e, and the magnetic field in Hc2 = c/2e 2 =  2Hc . The ring is placed in the plane x) y , the external magnetic field is directed along the z axis, and the indices 2D and 3D refer to two- and threedimensional operators, respectively. To solve the system of Eqs. (41) and (42), we generalized the approach of [15] for disks to our fat ring configuration. We apply a finite difference representation of the order parameter and the vector potential on a uniform Cartesian space grid (x) y), with a typical grid spacing of 0"15 and use the link variable approach [51] and an iteration procedure based on the Gauss–Seidel technique to find  . The vector potential is obtained with the fast Fourier transform  x=R )y=R = H0 x) −y /2 at the technique, where we set A S S boundary of a larger space grid (typically, RS = 4Ro ). To find the different vortex configurations, which include the metastable states, we search for the steady-state solutions of Eqs. (41) and (42) starting from different randomly generated initial conditions. Then we increase/decrease slowly the magnetic field and recalculate each time the exact vortex structure. We do this for each vortex configuration in a magnetic field range where the number of vortices stays the same. By comparing the dimensionless Gibbs free energies of the different vortex configurations

2

0

–2

(b) 0.0

0.5

1.0

1.5

2.0

H0/Hc2

Figure 22. (a) The free energy as a function of the applied magnetic field for a superconducting ring with Ro = 2"0, Ri = 0"4, d = 0"005, and  = 0"28. The different giant vortex states are shown by the thin solid curves and the multi-vortex states by the thick solid curves. (b) The magnetization for the same sample after averaging over the superconducting ring only. The different L states are given by the dashed curves, the ground state by the thin solid curves, and the multivortex states by the thick solid curves. The transitions from the multivortex state to giant vortex state are indicated by the open circles. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

(47)

where H0 is the applied magnetic field. H  is the magnetic field averaged over a sample/detector surface area S. As an example we take a fat ring with outer radius Ro = 4"0, thickness d = 0"005)  = 0"28, and different values of the inner radius. Figure 22a shows the free energy for such a ring with inner radius Ri = 0"4 as a function of the applied magnetic field H0 . The different giant vortex states are given by the thin solid curves and the multivortex state by the thick solid curves. The open circles indicate the transitions from the multivortex state to the giant vortex state. In such a ring, multivortex states exist with winding number L = 3 up to 7. For L = 3) 4, and 5 multivortices occur both as metastable states and in the ground state, while for L = 6 and 7 they are only found in the metastable state. Notice there is no discontinuity in the free energy at the transitions from the multivortex state to the giant vortex state for fixed winding number L. Figure 22b shows the magnetization M for this ring as a function of H0 after the field H is averaged over the superconducting ring (without the hole). The dashed curves give the results for the different L states, the thin solid curves are the results for the ground state, the thick solid curves are the results for the multivortex states,

and the open circles indicate the transition from the multivortex state to the giant vortex states. Notice that the latter transitions are smooth; there are no discontinuities in the magnetization. Now, we investigate the flux # through the hole for the L = 4 multivortex and giant vortex state for the case of the above ring. Figure 23 shows the flux # through a circular area of radius @ for different values of the applied magnetic field H0 . Curves 1, 2, 3, and 4 are the results for H0 /Hc2 = 0"72, 0"795, 0"87 (i.e., multivortex states), and 0"945 (i.e., giant vortex state), respectively. There is no qualitative difference between the four curves, that is, no qualitative difference between the multivortex states and the giant vortex state. In the inset, we show the flux through the hole with radius Ri = 0"4 and through the superconductor + hole as a function of the applied magnetic field for a fixed value of the winding number; that is, L = 4. The solid circles indicate the magnetic fields considered in the main figure and the open circle indicate the position of the transition from multivortex state to giant vortex state. The slope of the curves increases slightly at the applied magnetic field, where there is a transition form the multivortex state to the giant vortex state. This agrees with the result for a disk [16] that

603

Doubly Connected Mesoscopic Superconductors

8

4

flux through hole

10

0.16 0.14

4 2

0.10

1

7 6

–2 0.8

0.9

1.0

1.1

H0 /Hc2

Ro /ξ = 4.0 Ri /ξ = 0.4 2

0

+ superconductor

0.12 1

4

8

3 flux through hole

2

2

y/ξ

6

3

9

φ/φ0

φ/φ0

0.18

φ/φ0

4

11 0.20

–4

H0 /Hc2

d/ξ = 0.005

(1) 0.72

κ = 0.28

(2) 0.795

(b)

(c)

(d)

4

(3) 0.87

L=4

(a)

2

(4) 0.945

0

1

2

3

4

y/ξ

0 0

ρ/ξ –2

the giant ↔ multivortex transition is a second-order phase transition. The Cooper pair density F2 for the previous four configurations is shown in Figure 24. The darker the region, the larger the density and thus vortices are located at the white regions. At the magnetic field H0 /Hc2 = 0"72 (Fig. 24a) we clearly see three vortices. With increasing magnetic field, these vortices start to overlap and move to the center (Fig. 24b and c) and finally they combine to one giant vortex in the center (Fig. 24d). Please note that a theory based on the London limit will not be able to give such an intricate behavior, because in such a theory vortices are point-like objects. The free energies of the different vortex configurations were calculated for different values of the hole radius which we varied from Ri = 0 to Ri = 3"6. From these results we constructed an equilibrium vortex phase diagram. First, we assumed axial symmetry, where only giant vortex states  = F @ eiLC . occur and the order parameter is given by  @ In the phase diagram (Fig. 25) the solid curves separate the regions with different numbers of vortices (different L states). In the limit Ri → 0 we find the previous results of [16] for a superconducting disk. The radius of the giant vortex RL in the center of the disk increases with increasing L, because it has to accommodate more flux; that is,  RL / ∼ L/ H0 /Hc2 . Hence, if we make a little hole in the center of the disk, this will not influence the L → L + 1 transitions as long as Ri ≪ RL as is apparent from Figure 25. For a sufficiently large hole radius Ri , the hole starts to influence the giant vortex configuration and the magnetic field needed to induce the L → L + 1 transition increases. For example, the transition field from the L = 7 state to the

–4 –4

–2

0

2

4 –4

–2

0

x/ξ

2

4

x/ξ

Figure 24. The Cooper pair density corresponding to the four situations of Figure 23: H0 /Hc2 = 0"72 (a); H0 /Hc2 = 0"795 (b); H0 /Hc2 = 0"87 (c); and H0 /Hc2 = 0"945 (d). Dark regions indicate high density and light regions low density. The thick circles indicate the inner and the outer edge of the ring. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

L = 8 state reaches its maximum for a hole radius Ri ≈ 2"0 which occurs for H0 /Hc2 ≈ 1"5. The above rough estimate gives RL / ∼ 2"16, which is very close to RL / ∼ 2"0" Further increasing Ri , the hole becomes so large that more and more flux is trapped inside the hole, and consequently a φ/φ0 4.0

2

0

6

4

8

10

12

14

16

18

3.5 1

2

3

4

5

7

6

8

9

10

11

12

13

14

3.0

Superconducting state

2.5

Ri /ξ

Figure 23. The flux # through a circular area of radius @ for different values of the applied magnetic field, H0 . Curves 1, 2, 3, and 4 give the results for H0 /Hc2 = 0"72 (1), 0"795 (2), 0"87 (3) (i.e., multi-vortex states), and 0"945 (4) (i.e., giant vortex state), respectively. The inset shows the flux through the hole with radius Ri = 0"4 and through the superconductor + hole as a function of the applied magnetic field for a fixed value of the winding number, that is, L = 4. The solid circles indicate the magnetic fields considered in the main figure, and the open circles indicate the transition from the multivortex state to the giant vortex state. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

2.0

Normal state

1.5 1.0

Ro/ξ = 4.0

0.5

d/ξ = 0.005 κ = 0.28

L=0

0.0 0.0

0.5

1.0

1.5

2.0

H0 /Hc2

Figure 25. Equilibrium vortex phase diagram for a superconducting disk with radius Ro = 4"0 and thickness d = 0"005 with a hole with radius Ri in the center. The solid curves show the transitions between the different L states, the thick solid curve shows the superconducting/normal transition, and the dotted lines connect the results for Ri = 3"6 with the results in the limit Ri → Ro . The shaded regions indicate the multivortex states, and the dashed curves separate the multivortex states from the giant vortex states. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

604

circle is the inner radius of the ring. Low magnetic fields are given by light regions and darker regions indicate higher magnetic fields. In this way, multivortices in the superconducting area are dark spots. In Figure 26a the local magnetic field is shown for an applied magnetic field H0 /Hc2 = 0"895. Although the winding number is L = 4, there are only three vortices in the superconducting material and one vortex appears in the hole in the center of the disk. The phase difference is always given by > = L × 2, with L as the winding number. In Figure 26b a contour plot of the local magnetic field is shown for a ring with Ri = 1"0, which leads to L = 6 at H0 /Hc2 = 1"145. Only four vortices are in the superconducting material and one giant vortex is in the center (partially in the hole) with L = 2. Note that the flux # through the hole equals # ≈ 0"19#0 for the case of

4

y/ξ

2

0

–2

–4

(a)

4

2

y/ξ

smaller field is needed to induce the L → L + 1 transition. Because of our finite grid size, we were limited to Ri ≤ 3"6. The results we find for Ri = 3"6 are extrapolated to # = L + 1/2 #0 for Ri = Ro . The thick curve in Figure 25 gives the superconducting/normal transition. For low values of Ri , this critical magnetic field is independent of Ri , because the hole is smaller than the giant vortex state in the center and hence the hole does not influence the superconducting properties near the superconducting/normal transition. For Ri  2"0, this field starts to increase drastically. Therefore, more and more L states appear. In the limit Ri → Ro , the critical magnetic field is infinite and there are an infinite number of L states, which is a consequence of the enhancement of surface conductivity for very small superconducting samples [18]. Next, we consider the general situation where the order parameter is allowed to be a mixture of different giant vortex states and thus we no longer assume axial symmetry of the superconducting wavefunction. We found that the transitions between the different L states are not influenced by this generalization, but that for certain magnetic fields the ground state is given by the multivortex state instead of the giant vortex state. In Figure 25 the shaded regions correspond to the multivortex states and the dashed curves are the boundaries between the multivortex and the giant vortex states. For L = 1, the single vortex state and the giant vortex state are identical. In the limit Ri → 0, the previous results of [16] for a superconducting disk are recovered. For increasing hole radius Ri , the L = 2 multivortex state disappears as a ground state for Ri > 0"15. If Ri is further increased, the ground state for L = 5 up to L = 9 changes from giant vortex state to multivortex state and again to giant vortex state. For example, for Ri = 2"0 the multivortex state exists only in the L = 9 state. Notice that for small Ri the region of multivortex states increases and consequently the hole in the center of the disk stabilizes the multivortex states, at least for L > 2. For large Ri , that is, narrow rings, the giant vortex state is the energetic favorable one because confinement effects start to dominate, which impose the circular symmetry on F. For fixed hole radius Ri ≤ 2"0 and increasing magnetic field we find always at least one transition from giant vortex state to multivortex state and back to giant vortex state (re-entrant behavior). Note that the ground state for L ≥ 10 is in the giant vortex state irrespective of the value of the magnetic field. Near the superconducting/normal transition the superconducting ring is in the giant vortex state because now superconductivity exists only near the edge of the sample and consequently the superconducting state will have the same symmetry as the outer edge of the ring and thus it will be circular symmetric. Finally, for L ≥ 3 the multivortex state does not necessarily consist of L vortices in the superconducting material. Often it consists of a combination of a big vortex trapped in the hole and some multivortices in the superconducting material. This is clearly shown in Figure 26 where a contour plot of the local magnetic field is given for a superconducting disk with radius Ro = 4"0 and thickness d = 0"005 with a hole in the center with radius Ri = 0"6 (Fig. 26a) and Ri = 1"0 (Fig. 26b). As usual we took  = 0"28. The dashed thick circle is the outer radius, while the small solid black

Doubly Connected Mesoscopic Superconductors

0

–2

–4

(b) –4

–2

0

2

4

x/ξ

Figure 26. Contour plot of the local magnetic field for a superconducting disk with radius Ro = 4"0 and thickness d = 0"005  = 0"28 with a hole in the center with (a) radius Ri = 0"6 at H0 /Hc2 = 0"895 for L = 4 and (b) radius Ri = 1"0 at H0 /Hc2 = 1"145 for L = 6. The dashed thick circle is the outer radius; the small solid thick circle in the inner radius. Low magnetic fields are given by light regions and dark regions indicate higher magnetic fields. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

605

Doubly Connected Mesoscopic Superconductors

Figure 26a and # ≈ 0"36#0 for the case of Figure 26b and is thus not equal to a multiple of the flux quantum #0 .

8. ASYMMETRIC RINGS So far, we gave an overview of the influence of the size of the hole on the vortex configuration and the critical parameters for superconducting rings of different sizes. For small rings, only the axially symmetrical situation occurs, that is, the giant vortex states. For large rings the multivortex state can be stabilized for certain values of the magnetic field. In the next step, the axial symmetry is purposely broken by moving the hole away from the center of the superconducting disk over a distance a. As an example, we consider a superconducting disk with radius Ro = 2"0 and thickness d = 0"005 with a hole with radius Ri = 0"5 moved over a distance a = 0"6 in the x direction. Figure 27 shows the free energy and magnetization (defined through the field expelled from the (a) 0.0

Normal

–0.2

F/F0

–0.4

–0.6

–0.8

–1.0

5

(b) Ro/ξ = 2.0 Ri/ξ = 0.5

–M/Hc2 (10–4)

4

a/ξ = 0.6 3

d/ξ = 0.005 κ = 0.28

2 1 0 0.0

0.5

1.0

1.5

2.0

2.5

3.0

H0 /Hc2

Figure 27. The free energy (a) and magnetization after averaging over the superconducting ring only (b) as function of the applied magnetic field for a superconducting disk with radius Ro = 2"0 and thickness d = 0"005 with a hole with radius Ri = 0"5 moved over a distance a = 0"6 in the x direction. The solid curve indicates the ground state; the dashed curve indicates the results for increasing and decreasing field. The vertically dotted lines separate the regions with different vorticity L. The insets show the Cooper pair density at magnetic field H0 /Hc2 = 0"145, 1"02, 2"145, and 2"52, where the ground state is given by a state with L = 0) 1) 2, and 3, respectively. High Cooper pair density is given by dark regions, while light regions indicate low Cooper pair density. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

superconducting ring without the hole) as a function of the magnetic field. The solid curve indicates the ground state, while the dashed curves indicate the metastable states for increasing and decreasing fields. The vertically dotted lines separate the regions with different winding number L. Notice that hysteresis is only found for the first transition from the L = 0 to the L = 1 state and not for the higher transitions, which are continuous. At the transition from the L = 1 state to the L = 2 state and further to the L = 3 state, the free energy and the magnetization vary smoothly (see Fig. 27a and b). In the insets of Fig. 27a, we show the Cooper pair density for such a sample at the magnetic fields: H0 /Hc2 = 0"145, 1"02) 2"145, and 2"52, where the ground state is given by a state with L = 0) 1) 2, and 3, respectively. High Cooper pair density is indicated by dark regions, whereas light regions indicate low Cooper pair density. For H0 /Hc2 = 0"145, we find a high Cooper pair density in the entire superconducting ring. There is almost no flux trapped in the circular area with radius smaller than Ro . After the first transition at H0 /Hc2 ≈ 0"75, suddenly more flux is trapped in the hole, which substantially lowers the Cooper pair density in the superconductor. Notice that the trapped flux tries to restore the circular symmetry in the Cooper pair density and that the density of the superconducting condensate is largest in the narrowest region of the superconductor. The next inset shows the Cooper pair density of the L = 2 state where an additional vortex appears. Some flux is passing through the hole (i.e., winding number equals 1 around the hole), while one flux line is passing through the superconducting ring and a local vortex (the normal region with zero Cooper pair density) is created. In the L = 3 state superconductivity is destroyed in part of the sample which contains flux with winding number L = 2 and the rest of the flux passes through the hole. Hence, by breaking the circular symmetry of the system, multivortex states are stabilized. Remember that for the corresponding symmetric system, that is, a = 0, only giant vortex states were found. Having the magnetic fields for the different L → L + 1 transitions for superconducting rings with different positions of the hole, that is, different values of a, we constructed the phase diagram shown in Figure 28. The thin curves (solid curves when the magnetization is discontinuous and dashed curves when the magnetization is continuous) indicate the magnetic field at which the transition from the L state to the L + 1 state occurs, while the thick solid curve gives the superconducting/normal transition. To show that the stabilization of the multivortex state due to an off-center hole is not peculiar for Ro = 2"0, we repeated the previous calculation for a larger superconducting disk with radius Ro = 5"0 and thickness d = 0"005 containing a hole with radius Ri = 2"0. In Figure 29a the Cooper pair density is shown for such a system with the hole in the center, while in Figure 29b the hole is moved away from the center over a distance a = 1"0 in the negative y direction. The externally applied magnetic field is the same in both cases, H0 /Hc2 = 0"77, which gives a winding number L = 4 and L = 5 for the ground state of the symmetric and the nonsymmetric geometry, respectively. The assignment of the winding number can be easily checked from Figure 29c and d which shows contour plots of the

606

Doubly Connected Mesoscopic Superconductors

corresponding phase of the superconducting wavefunction. If the hole is at the center of the disk, the ground state is a giant vortex state. If the center of the hole is moved to the position x/) y/ = 0) −1 , two vortices appear in the superconducting material while the hole contains three vortices. Notice that in this case, although the magnetic field is kept the same and the amount of superconducting material is not altered, changing the symmetry of the system alters the winding number.

1.4

1.2

L=0

1

3

2

1.0

a/ξ

0.8

Superconducting

0.6

State Normal

0.4

9. SADDLE POINT STATES

State 0.2 0.0 0.0

1.0

0.5

2.0

1.5

3.0

2.5

3.5

H0 /Hc2

Figure 28. The relation between the displacement a and the transition magnetic fields for a superconducting ring with R0 / = 2"0) d/ = 0"005, Ri / = 0"5, and  = 0"28. The thin solid curves give the L → L + 1 transition when it corresponds to a first-order phase transition while a dashed curve is used when it is continuous. The thick solid curve is the superconducting/normal transition. Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

6

a/ξ = 0.0 4

y/ξ

2 0

Up to now we only discussed stable vortex configurations, which correspond to minima in the free energy landscape. These minima are separated from each other by saddle points (Fig. 30). Therefore, the states corresponding to the saddle points (i.e., the saddle point states) describe the transition from one vortex state to another. In other words, by studying the saddle point states, we can investigate the flux entree and expulsion in mesoscopic rings; that is, how does the vortex enter or leave the superconductor? We considered superconducting rings with inner radius Ri , outer radius Ro , and thickness d [52]. These mesoscopic superconducting systems are immersed in an insulating medium in the presence of a perpendicular uniform magnetic field H0 . To solve this problem, we follow the numerical approach of Schweigert and Peeters [8]. For very thin rings, that is, Wd ≪ 2 , with W = Ro − Ri as the width of the ring, the demagnetization effects can be neglected and the Ginzburg–Landau functional can be written as     G = Gn + d r  2 +  4 +  ∗  L (48) 2

–2 –4

(a)

–6 6

(c)

a/ξ = 1.0 free energy

4

y/ξ

2 0 –2 –4

(b)

–6 –6

–4

–2

0

x/ξ

2

4

6

(d) –6

–4

–2

0

2

4

6

x/ξ

Figure 29. The Cooper pair density for a superconducting disk with radius Ro = 5"0 and thickness d = 0"005 with a hole with radius Ri = 2"0 (a) in the center and (b) moved away form the center over a distance a = 1"0 in the negative y direction. The applied magnetic field is the same in both cases: H0 /Hc2 = 0"77. High Cooper pair density is given by dark regions and low Cooper pair density by light regions. The corresponding contour plot of the phase of the superconducting wavefunction is given in (c) and (d). Reprinted with permission from [39], B. J. Baelus et al., Phys. Rev. B 61, 9734 (2000). © 2000, American Physical Society.

path

Figure 30. Schematical view of the free energy in functional space depicting two minima with L = 2 and L = 3 and the saddle point connecting them. The Cooper pair densities of these three states are shown in the insets. High (low) Cooper pair density is given by dark (light). Reprinted with permission from [8], V. A. Schweigert and F. M. Peeters, Phys. Rev. Lett. 83, 2409 (1999). © 1999, American Physical Society.

607

Doubly Connected Mesoscopic Superconductors

where G and Gn are the free energies of the superconducting and the normal states,  is the complex order parameter, and  and  are the GL coefficients which depend on the sample temperature.  L is the kinetic energy operator for Cooper pairs of charge e∗ = 2e and mass m∗ = 2m; that is, ∗   L = −i  − e∗ A/c /2m

(49)

 = e# H0 @/2 is the vector potential of the uniform where A @ and #. magnetic field H0 written in cylindrical coordinates  By expanding the order parameter  = i Ci i in the orthonormal eigenfunctions of the kinetic energy operator  Li = 5i i [5, 6, 15, 16], the difference between the superconducting and the normal state Gibbs free energy can be written in terms of complex variables as  ij ∗ ∗ A C C C C (50) 2 kl i j k l ij  i∗ j∗ k l are calwhere the matrix elements Akl = d r culated numerically. The boundary condition for these i , corresponding to zero current density in the insulator media, is given by Eq. (15). These eigenenergies 5i and the eigenfunctions i depend on the sample geometry. For thin axial symmetric samples the eigenfunctions have the form j= n)l @) # = exp il# fn @ , where l is the angular momentum and the index n counts different states with the same l and equals the number of nodes in the radial direction. Thus, the order parameter  can be written as  Cn)l n)l (51) F= F = G − Gn =  + 5i Ci Ci∗ +

n

Starting from a randomly chosen initial set of coefficients, we calculate a nearby minimum of the free energy by moving in the direction of the negative free energy gradient −Lk . The set of coefficients of this minimum then determines the ground state or a metastable state. Starting from the initial set of coefficients we can also calculate a nearby saddle point state by moving to a minimum of the free energy in all directions, except the one that has the lowest eigenvalue. In this direction we move to a local maximum. Repeating this procedure for many randomly chosen initial sets of coefficients HCi I for fixed magnetic field, we find the different possible superconducting states and saddle point states. To calculate the magnetic field dependence we start from a superconducting state at a certain field and we change the applied field by small increments. By moving into the direction of the nearest minimum or saddle point, the corresponding state will be found for the new magnetic field, provided that the field step is small enough. Just like for the discussion of the stable states, we also make a distinction between small and large rings for the discussion of the saddle point states. As mentioned above, in small rings only giant vortex states are stable, while in larger rings it is possible that multivortex states are stabilized.

l

We do not restrict ourselves to the lowest Landau level approximation (i.e., n = 1) and expand the order parameter over all eigenfunctions with energy 5i < 5∗ , where the cutting parameter 5∗ is chosen such that increasing it does not influence the results. The typical number of complex components used is in the range N = 30–50. Thus the superconducting state is mapped into a 2D cluster of N particles with coordinates xi ) yi ↔ Re Ci ) Im Ci , whose energy is determined by the Hamiltonian (50). The energy landscape in this 2N + 1 dimensional space is studied where the local minima and the saddle points between them will be determined together with the corresponding vortex states. To find the superconducting states and the saddle point states we use the technique described in [8]. A particular state is given by its set of coefficients HCi I. We calculate the free energy in the vicinity of this point G = G C n − G C where HC n I is the set of coefficients of a state very close to the initial one. This free energy is expanded to second order in the deviations  = C n − C, G = Fm ∗m + Bmn n ∗m + Dmn ∗n ∗m + c"c"

(52)

where mj

Fm =  + 5i Cm + Akl Cj Ck∗ Cl Bmn =  + 5m Imn + Dmn =

and Imn is the unit matrix. Using normal coordinates m = k xk Qm we can rewrite the quadratic form as G = 2 Lk xk + ?k xk2 . To find the eigenvalues ?k and the eigenvectors Lk we solve numerically the equation        B + Re D  Re Qk  Im D   Re Qk       = ?k  "   Im D  Im Qk  B − Re D   Im Qk  (56)

∗ 2Amn kl Ck Cl

Amn kl Ck Cl

(53) (54) (55)

9.1. Small Rings: Giant Vortex State As an example, we consider a superconducting ring with radius Ro = 2"0 and hole radius Ri = 1"0. In Figure 31 the free energy is shown as a function of the applied magnetic field for the different L states (solid curves) together with the saddle point states (dashed curves). We find giant vortex states with L = 0) 1) 2) 3, and 4. Comparing this result with the result for a disk with radius R = 2"0, we find that more L states are possible and the superconducting/normal transition moves to larger magnetic fields [39]. The inset shows the energy barrier U for the transitions between the different L states as a function of the difference between the applied magnetic field H0 and the L → L + 1 transition field HL→L+1 . For increasing L, the height of the energy barrier and the difference between the penetration and the expulsion field decreases. The energy barrier near its maximum can be approximated by U /F0 = Umax /F0 +  H − Hmax /Hc2 ) and we determined the slope L→L+1 ; 0→1 = −0"8 for H  Hmax and 0"9 for H  Hmax , 1→2 = −0"6 for H  Hmax and 0"75 for H  Hmax , 2→3 = −0"3 for H  Hmax and 0"42 for H  Hmax , and 3→4 = −0"013 for H  Hmax and 0"038 for H  Hmax . The slope decreases again for increasing L and the absolute value of the slope for H  Hmax is smaller than that for H  Hmax for every L. Next, we investigate the 2 ↔ 3 saddle point. At H0 /Hc2 = 2"01 (expulsion field) and 2"535 (penetration field) the saddle point state equals the giant vortex states with L = 3 and L = 2, respectively. The transition between these two

608

Doubly Connected Mesoscopic Superconductors 2

0.0

4

Ro /ξ = 2.0 3

Ri /ξ = 1.0 0.20

1

L=0

y/ξ

–0.4 UL↔L+1/F0

F/F0

0.15

2

–0.6

× 10

2

(a)

0.00 –0.4

L=0

–0.2

0.0

0.2

1

2.0

2.5

3.0

3.5

H0 /Hc2

Figure 31. The free energy for a superconducting ring with radius Ro = 2"0 and hole radius Ri = 1"0 as a function of the applied magnetic field for the different giant vortex states (solid curves) and for the saddle point states (dashed curves). The inset shows the energy barrier U for the transitions between different L states as a function of the difference between the applied magnetic field H0 and the L → L + 1 transition field HL→L+1 . Reprinted with permission from [52], B. J. Baelus et al., Phys. Rev. B 63, 144517 (2001). © 2001, American Physical Society.

giant vortex states is illustrated in Figure 32a–d which shows the spatial distribution of the superconducting electron density  2 corresponding with the open circles in the inset of Figure 31 at H0 /Hc2 = 2"1, 2"2, 2"315 (i.e., the barrier maximum), and 2"4, respectively. High (low) density is given by dark (light) regions. With increasing field one vortex moves from inside the ring, through the superconducting material, to outside the ring. From Figure 32c one may infer that the Cooper pair density is zero along a radial line and that the vortex is, in fact, a sort of line. That this is not the case can be seen from the left inset of Figure 33, which shows the Cooper pair density F2 along this radial line for H0 /Hc2 = 2"315. The Cooper pair density in the superconducting material is zero only at the center of the vortex, which is situated at xmin / ≈ 1"5, and  2 is very small otherwise; that is,  2 < 0"01. In Figure 33 the position of the vortex (i.e., of xmin ) is shown as a function of the applied field. Over a narrow field region the vortex moves from the inner boundary toward the outer boundary. From H0 /Hc2 = 2"01–2"25 the center of the vortex is still situated in the hole, but the vortex already influences the superconducting state (see, e.g., Fig. 32a and b). From H0 /Hc2 = 2"36–2"535 the center of the vortex lies outside the ring, but it still has an influence on the saddle point (see, e.g., Fig. 32d). In the region H0 /Hc2 = 2"25–2.36 the center of the vortex is situated inside the superconductor. This is also illustrated by the contour plot (right inset of Fig. 33) for the phase of the order parameter at H0 /Hc2 = 2"315, corresponding with the open circle in Figure 33. When encircling the superconductor near the inner boundary of the ring, we find that the phase difference > is equal to 2 × 2 which implies vorticity L = 2. When encircling the superconductor near the outer boundary, we find vorticity L = 3. If we choose a path around the vortex (located at xmin ), the phase changes with 2 and thus L = 1. At the transition field H0 /Hc2 = 2"315 the center of the vortex of the saddle point is clearly not

0

H0 /Hc2 = 2.315

H0 /Hc2 = 2.4

–1

(c)

(d)

–2 –2

–1

0

1

–2

–1

0

1

2

x/ξ

x/ξ

Figure 32. The spatial distribution of the superconducting electron density  2 of the transition between the giant vortex states with L = 2 and L = 3 for a superconducting ring with Ro = 2"0 and Ri = 1"0 at H0 /Hc2 = 2"1 (a), 2"2 (b), 2"315 (c), and 2"4 (d). High density is given by dark regions and low density by light regions. Reprinted with permission from [52], B. J. Baelus et al., Phys. Rev. B 63, 144517 (2001). © 2001, American Physical Society.

situated at the outer boundary as was the case for superconducting disks (see e.g., [8, 52]).

9.2. Large Rings: Multivortex States First, we consider superconducting rings with radius Ro = 4"0 and hole radius Ri = 1"0. In Figure 34 the free energy is shown as a function of the applied magnetic field. The different L states are given by solid curves for giant vortex 1.0

2.0

1.8

L=2

1.6

xmin/ξ

1.5

y/ξ

(H0-HL→L+1)/Hc2 1.0

(b)

–2 3

1

0.5

H0 /Hc2 = 2.2

–1

0.10

0.05

0.0

H0 /Hc2 = 2.1

1

–0.8

–1.0

0

|ψ|2 (10–2)

–0.2

0.8

H0/Hc2 =

0.6

2.315

xmin

0.4 0.2 0.0 0.0

1.4

0.5

1.0

1.5

2.0

x/ξ

1.2 Ro/ξ = 2.0 L=3

1.0 2.0

Ri/ξ = 1.0 2.1

2.2

2.3

2.4

2.5

H0 /Hc2

Figure 33. The radial position of the vortex line in the saddle point for the 2 ↔ 3 transition through the superconductor with radius Ro = 2"0 and Ri = 1"0. The left inset shows the Cooper pair density along the x-direction at H0 /Hc2 = 2"315, and the right inset is a contour plot of the phase of the order parameter at H0 /Hc2 = 2"315. Reprinted with permission from [52], B. J. Baelus et al., Phys. Rev. B 63, 144517 (2001). © 2001, American Physical Society.

609

Doubly Connected Mesoscopic Superconductors

0.0

Ro/ξ = 4.0 Ri/ξ = 1.0

6

–0.2

7

8

9

10

5 4 3

F/F0

–0.4

(a)

2

–0.6

–0.8

1

–1.0

(b)

L=0

0.0

0.5

1.0

(c) 1.5

H0/Hc2

Figure 34. The free energy for a superconducting ring with Ro = 4"0 and Ri = 1"0 as a function of the applied magnetic field for the different L states (solid curves for giant vortex states and dashed curves for multivortex states) and the saddle point states (dotted curves). The open circles correspond to the transition between the multivortex state and the giant vortex state for fixed L. The inset shows the spatial distribution of the superconducting electron density  2 for the multivortex state with L = 4 at H0 /Hc2 = 0"8 (a), L = 5 at H0 /Hc2 = 0"9 (b), and L = 6 at H0 /Hc2 = 1"0 (c). High Cooper pair density is given by dark regions; low Cooper pair density by light regions. Reprinted with permission from [52], B. J. Baelus et al., Phys. Rev. B 63, 144517 (2001). © 2001, American Physical Society.

L=0

0.14

Ri/ξ = 3.0

Umax/F0

0.12

0.12 0.10

2.0 0.08

0.04

0.0 1.0

0.00

0.08 0.06

1

1.5

Htr/Hc2

UL↔L+1/F0

states and dashed curves for multivortex states, and the saddle point states are given by the dotted curves. The open circles correspond to the transition between the multivortex state and the giant vortex state for fixed L. These transitions occur at HMG /Hc2 = 0"93, 1"035, and 1"14 for L = 4, 5, and 6, respectively. Notice that for such a small hole in the disk the maximum number of L (i.e., L = 10) is the same as for the disk case without a hole. The spatial distribution of the superconducting electron density  2 is depicted in the insets (a–c) of Figure 34 for the multivortex state with L = 4 at H0 /Hc2 = 0"8, L = 5 at H0 /Hc2 = 0"9, and L = 6 at H0 /Hc2 = 1"0, respectively. High (low) Cooper pair density is given by dark (light) regions. Notice that there are always L − 1 vortices in the superconducting material and one vortex appears in the hole, that is, in the center of the ring. The energy barriers for the transitions between the different L states are shown in Figure 35 as a function of the applied magnetic field. By comparing this with the energy barriers for a disk with no hole [52], we see that the barrier heights and the transition fields are very different. Therefore we show in the insets of Figure 35 the maximum height of the energy barrier Umax and the L ↔ L + 1 transition field Htr as a function of L for superconducting disks with no hole (squares) and with a hole of radius Ri = 1"0 (circles), 2"0 (triangles), and 3"0 (stars). In all cases the height of the energy barrier decreases and the transition fields increase with increasing L. By comparing the situation with no hole and that with a small hole with Ri = 1"0, we see that the barriers for L ≤ 1 are higher for the disk with a hole with Ri = 1"0 than for Ri = 0"0, whereas they are smaller when L > 1. Notice also that the value of the L → L + 1 transition field is sensitive to the presence of the hole with radius

1.0 0.5

0.04

Ro/ξ = 4.0

0.0 0

2

0.02

3

1

2

3

4

5

6

7

8

9

L 4

5 6

7

8

9

0.00 0.0

1.0

0.5

1.5

H0/Hc2

Figure 35. The energy barrier U for the transitions between the different L states in a superconducting ring with Ro = 4"0 and Ri = 1"0 as a function of the applied magnetic field. The insets show the maximum height of the energy barrier Umax and the transition field Htr as a function of L for rings with Ro = 4"0 and Ri = 0"0, 1"0, 2"0, and 3"0. Reprinted with permission from [52], B. J. Baelus et al., Phys. Rev. B 63, 144517 (2001). © 2001, American Physical Society.

Ri = 1"0 for small L and insensitive for larger L. The reason is that for small L > 0 such a central hole always has one vortex localized inside, which favors certain vortex configurations above others, whereas for larger L in both cases only giant vortices appear with sizes larger than the hole size and the presence of the hole no longer matters. For larger holes the energy barrier decreases more slowly, because the free energy of the different L states shows a more parabolical type of behavior as a function of the magnetic field. The transition field has a much smaller dependence on the radius of the hole. Notice that the transition field for Ri = 3"0 is linear as a function of L for small holes and L ≤ 9. This is in good agreement with the results in the narrow ring limit, where the transition between states with different vorticity L occurs when the enclosed flux # equals L + 1/2 #o [48]. Next, we investigate the saddle point states in these superconducting rings. We make a distinction between different kinds of saddle point states: (1) between two giant vortex states, (2) between a multivortex and a giant vortex state, (3) between two multivortex states with the same vorticity in the hole and different vorticity in the superconducting material, and (4) between two multivortex states with the same vorticity in the superconducting material but different vorticity in the hole. The first saddle point transition was already described for the case of small superconducting rings (see Figs. 32 and 33). Next, we study the saddle point state between a multivortex state with L = 5 and a giant vortex state with L = 4 for the previously considered ring with radius Ro = 4"0 and hole radius Ri = 1"0. Figure 36a–f shows the Cooper pair density for these saddle point states at H0 /Hc2 = 0"83, 0"88, 0"93, 0"965 (i.e., the barrier maximum), 1"03, and 1"06, respectively. High (low) Cooper pair density is given by dark (light) regions. For increasing field one vortex moves to the outer boundary, while the others move to the center of the ring where they create a giant vortex state. Note that the giant vortex state is larger than

610

Doubly Connected Mesoscopic Superconductors 4 –0.20 L=9

0

–0.25

F/F0

y/ξ

2

–2

(a)

–0.30

(b)

–4

2

–0.35

Ro/ξ = 6.0 L=8

y/ξ

Ri/ξ = 2.0 0

0.6

0.7

0.8

H0 /Hc2 –2

(c)

(d)

(e)

(f)

Figure 37. The free energy of the multivortex states with L = 8 and L = 9 (solid curves) and the saddle point state (dashed curves) between these multivortex states for a superconducting ring with Ro = 6"0 and Ri = 2"0 as a function of the applied magnetic field. The lower insets show the spatial distribution of the superconducting electron density  2 at the transition field H0 /Hc2 = 0"695 for L = 8 and L = 9. The upper insets show the spatial distribution of the superconducting electron density  2 for the saddle point states indicated by the open circles, that is, at H0 /Hc2 = 0"63 (a), 0"695 (b), and 0"76 (c). High Cooper pair density is given by dark regions and low Cooper pair density by light regions. Reprinted with permission from [52], B. J. Baelus et al., Phys. Rev. B 63, 144517 (2001). © 2001, American Physical Society.

–4

y/ξ

2

0

–2

–4 –4

–2

0

x/ξ

2

–4

–2

0

2

4

x/ξ

Figure 36. The Cooper pair density for the saddle point state transition between a multivortex state with L = 5 and a giant vortex state with L = 4 at H0 /Hc2 = 0"83 (a), 0"88 (b), 0"93 (c), 0"965 (d), 1"03 (e), and 1"06 (f). High Cooper pair density is given by dark regions and low Cooper pair density by light regions. Reprinted with permission from [52], B. J. Baelus et al., Phys. Rev. B 63, 144517 (2001). © 2001, American Physical Society.

the hole and therefore it is partially situated in the superconductor itself. To study saddle point transitions between different multivortex states we have to increase the radius of the ring to favor the multivortex states. Therefore, we consider a ring with radius Ro = 6"0 and hole radius Ri = 2"0. Figure 37 shows the free energy of multivortex states with L = 8 and L = 9. In both cases three vortices are trapped in the hole. The lower insets show the spatial distribution of the superconducting electron density  2 at the transition field H0 /Hc2 = 0"695 for L = 8 and L = 9. It is clear that there are only five and six vortices in the superconducting material, respectively. The free energy of these multivortex states is shown by solid curves, whereas the saddle point energy between these states is given by the dashed curve. Notice further that there is no transition from the multivortex states to the giant vortex states with L = 8 and 9 as long as these states are stable. The spatial distribution of the superconducting electron density  2 for this saddle point state is depicted in the upper insets at the magnetic fields H0 /Hc2 = 0"63, 0"695 (the barrier maximum), and 0"76, respectively.

For increasing field one vortex moves from the superconducting material to the outer boundary and hence the vorticity changes from L = 9 to L = 8" Notice that the vorticity of the interior boundary of the ring does not change. The fourth type of saddle point state to discuss is the L → L + 1 transition between two multivortex states with the same vorticity in the superconducting material but with a different vorticity in the hole. For Ro / = 4 and Ro / = 6 we did not find such transitions regardless of the hole radius. This means that at least for these radii there is no transition between such states that describes the motion of one vortex from the hole through the superconducting material toward the outer insulator. Finally, we investigated the influence of the hole radius on the barrier for a fixed outer ring radius. Figure 38a shows the maximum barrier height, that is, the barrier height at the thermodynamic equilibrium L → L + 1 transition, as a function of the hole radius Ri for a ring with radius Ro = 4"0 for the transition between the Meissner state and the L = 1 state (solid curve) and for the transition between the L = 1 state and the L = 2 state (dashed curve). For increasing hole radius, the barrier height of the first transition rapidly increases in the range Ri = 0"1 to Ri = 1"5 and decreases slowly afterward. For a superconducting disk with radius Ro = 4"0 with a hole in the center with radius Ri = 1"5 the maximum barrier height for the 0 → 1 transition is twice as large as that for a superconducting disk without a hole. The barrier height of the second transition first decreases, then rapidly increases in the range Ri = 0"6 to Ri = 2"5 and then slowly decreases again. In this case the maximum barrier height for a superconducting disk with

611

Doubly Connected Mesoscopic Superconductors

Umax/F0

0.11 0.09

L=0→L=1 L=1→L=2

0.07 0.05

Ro/ξ = 4.0

0.03

(b) 0.6

H/Hc2

0.4

Hp

0.2 Htr 0.0 He –0.2 0

1

2

3

4

Ri /ξ

Figure 38. (a) The maximum barrier height as a function of the hole radius Ri for a ring with radius Ro = 4"0 for the transition between the Meissner state and the L = 1 state (solid curve) and the transition between the L = 1 state and the L = 2 state (dashed curve) and (b) the transition magnetic field Htr , the expulsion magnetic field He , and the penetration magnetic field Hp as a function of the hole radius Ri for the transition between the Meissner state and the L = 1 state (solid curve) and the transition between the L = 1 state and the L = 2 state (dashed curve). Reprinted with permission from [52], B. J. Baelus et al., Phys. Rev. B 63, 144517 (2001). © 2001, American Physical Society.

a hole with radius Ri = 2"5 is three times as large as that for a superconducting disk without a hole. Hence, changing the hole radius strongly influences the maximum height of the barrier. In Figure 38b we plot the characteristic magnetic fields of the barrier as a function of the hole radius (i.e. the transition magnetic field Htr ), the expulsion magnetic field He , and the penetration magnetic field Hp for the 0 → 1 transition by solid curves and for the 1 → 2 transition by the dashed curve. For the 0 → 1 transition the characteristic magnetic fields decrease with increasing hole radius. For the 1 → 2 transition the characteristic magnetic fields first increase to a maximum and then decrease. This behavior was already encountered in Figure 25. Notice that the position of the minimum in Umax coincides with the position of the maximum in Htr .

10. EXPERIMENTAL RESULTS Two different types of experiments have been performed on single mesoscopic rings: resistivity measurements and magnetization measurements.

10.1. Resistivity Measurements Moshchalkov et al. [19] performed resistivity measurements on a square Al dot and on a square Al ring with connected leads (see the insets of Fig. 39). The sample geometry was 160 1 µm

0.15 µm

0.13

To measure the resistivity, leads are attached to the sample and a transport current is applied through the leads. By measuring the resistivity of the non-fully superconducting state one can obtain information about the superconducting/normal transition. Most of the time this information is plotted in a H − T phase diagram, which shows transitions between vortex states with different vorticity L. With this technique it is possible to investigate the dependence of the critical magnetic field and temperature, but it does not provide any information about the real vortex structure. Another restriction is that the leads can influence the result. Hall magnetometry is revealing itself as a powerful tool for obtaining information on a single mesoscopic superconductor away from the superconducting/normal phase boundary, that is, deep inside the superconducting state. Thin superconducting samples are placed on top of a Hall probe created in a two-dimensional electron gas. In the ballistic regime the Hall resistance is directly proportional to the average of the magnetic field through the junction [44]. The   in the junction averaged value of the magnetic field H area is determined by the expulsion of the field by the superconductor, which is placed on top of the junction. Because the magnetization is given by the difference between the averaged field and the applied field, one can obtain the magnetization of single superconductors and compare them with the theoretical values (see, e.g., [13, 14]). By using this technique, it has been possible to show that a single mesoscopic superconductor in a magnetic field does not exhibit the standard magnetization dependencies of bulk superconductors.

120

80 "Hc1"

40

1 µm

0.15 µm

(a)

H (G)

0.15

0 0.85

0.90

0.95

1.00

T/Tc

Figure 39. The measured superconducting phase boundary in the H − T /Tc plane for a 1 ;m×1 ;m Al square ring (square symbols) and by a 1 ;m×1 ;m filled Al square (circular symbols). The solid lines correspond to calculations based on the linear Ginzburg–Landau theory. The dashed line shows the phase boundary for a straight wire. Reprinted with permission from [19], V. V. Moshchalkov et al., Nature 373, 319 (1995). © 1995, Macmillan Magazines Ltd.

612

Doubly Connected Mesoscopic Superconductors

where Ri and Ro are the inner and outer boundaries of the ring, respectively. The first term gives a monotomic shift of Tc H . The second term describes the oscillatory Tc H variation caused by the change of orbital quantum number L with increasing magnetic flux. Using the actual sample dimensions the theoretical phase boundary was calculated (see the solid lines in Fig. 39), which agrees well with the experimental data.

10.2. Magnetization Measurements Pedersen et al. [3] performed magnetization measurements of a micron-sized superconducting aluminium loop placed on top of a ;-Hall magnetometer. The ;-Hall magnetometer was etched out of a GaAs/Ga0"7 Al0"3 As heterostructure. A symmetrical 4 ;m × 4 ;m Hall geometry was defined by standard electron beam lithography on top of the heterostructure. In a later processing step a lift-off mask was defined on top of the ;-Hall probe by electron beam lithography. After deposition of a d = 90-nm-thick layer of aluminium and lift-off the sample looked as shown in Figure 40. The mean radius of the aluminium loop is R = 2"16 ;m and the average wire width w is 316 ± 40 nm. The superconducting coherence length was estimated to be approximately o = 180 nm, corresponding to a bulk critical field of Hc2 = -0 /2o2 ≈ 100 G. By using the expression n-0 = n

h = > ;0 H R2 2e

(58)

NBI

1 µm x 15 , 000

2 . 5KV

7 mm

Figure 40. Scanning electron microscope image of a ;-Hall probe, the cross-section of the etched ;-Hall probe is 4 × 4 ;m2 . The mean radius of the superconducting aluminium loop deposited on top of the ;-Hall magnetometer is 2"16 ;m, and the difference between the outer and inner radius is 314 nm. Reprinted with permission from [3], S. Pedersen et al., Phys. Rev. B 64, 104522 (2001). © 2001, American Physical Society.

where A = R2 is the area of the loop given by its mean radius R, it is found that a single flux jump (n = 1) corresponds to a magnetic field periodicity given by > ;0 H = 1"412 G for the ring shown in Figure 40. In Figure 41 the measured local magnetization ;0 M detected by the ;-Hall is shown as a function of magnetic field intensity ;0 H . The measurement was performed at T = 0"36 K on the device of Figure 40. The curve shows a series of distinct jumps corresponding to abrupt changes in the magnetization of the superconducting loop. The difference in the magnetic field intensity between two successive flux jumps is approximately given by > ;0 H = 1"4 G or > ;0 H = 2"8 G which corresponds to either single or double flux jumps (n = 1 or n = 2). Large flux jumps n > 1 or flux avalanches

0.8

–µ0M [Gauss]

defined by combining electron beam lithography and lift-off techniques. The resistance of the 25-nm-thick Al samples was measured using a four-probe technique with an alternating current resistance bridge. The normal state resistivity @ ≈ 4"5 × 10−6 M cm of the Al structures as well as the resistance ratio R 300 K /R 4"2 K ≈ 2"1 indicate a pronounced metallic character. Detailed atomic force microscopy and scanning electron microscopy studies of the Al films confirm the presence of a smooth and continuous film surface. The normal superconductivity phase boundary was constructed in the H − T phase diagram by measuring the temperature shift of the midpoint of the normal-tosuperconducting resistive transition at different magnetic fields. The latter was always applied perpendicular to the plane of the sample. The square symbols in Figure 39 give the superconducting phase boundary in the H − T /Tc plane for the 1 ;m × 1 ;m Al square ring and the circular symbols for the 1 ;m × 1 ;m filled Al square. The solid lines in Figure 39 correspond to calculations based on the linear Ginzburg–Landau theory. The dashed line shows the phase boundary for a straight wire. The phase boundary for the square ring shows classic Little–Parks oscillations, which are related to fluxoid quantization in the ring. The phase boundary for the ring can be obtained by averaging the one-dimensional result [31] over the different radii imposed by the finite linewidth of the ring [48]. Such a calculation results in   Tc H − Tc H = 0  0 wH 2 = √ Tc H = 0 3-0   2 HRi Ro 2  0 ) (57) L− = Ri Ro -0

0.4

0.0

–0.4

–0.8 –100

–50

0

50

100

µ0H [Gauss]

Figure 41. Measured magnetization ;0 M detected by the ;-Hall probe as a function of magnetic field intensity ;0 H of the device presented in Figure 40. The curve displays distinct jumps corresponding to the abrupt changes in magnetization of the superconducting loop when the system changes state. The measurements were performed at T = 0"36 K. Reprinted with permission from [3], S. Pedersen et al., Phys. Rev. B 64, 104522 (2001). © 2001, American Physical Society.

613

Doubly Connected Mesoscopic Superconductors

occur when the system is trapped in a metastable state. It was generally observed that these flux avalanches become more pronounced with decreasing temperature, at low magnetic field intensities, and for wide loops. Furthermore the flux avalanches were sensitive to the cooling procedure. The energy barrier causing the metastability of the eigenstates of the loop is due to either the Bean–Livingston surface barrier or the volume barrier, or even an interplay of both [53–55]. In Figure 42 the magnetic field intensity difference between successive jumps > ;0 H is given as a function of the magnetic field intensity. One can see that the magnetic field intensity difference between the observed jumps is, to a high degree of accuracy, given by integer values of 1"412 G, that is, the flux quantum. For magnetic fields lower than 40 G double flux jumps dominate, whereas at higher magnetic fields only single flux jumps are observed. The figure shows both an up-sweep and a down-sweep as indicated by the arrows. Pedersen et al. [3] also studied thicker loops where the flux avalanches are much more pronounced. In Figure 42 it is seen that a small systematic variation of the value of the flux jumps occurs when the magnetic field intensity is changed. This fine structure appears as a memory effect in the sense that with increasing (decreasing) magnetic field intensity the size of the flux jumps decreases (increases). Thus these deviations depend not only on the size of the magnetic field intensity, but also on which direction the magnetic field intensity was swept during measurements. The data presented in Figure 42 have been replotted in Figure 43 in the following way. Eq. (58) is used to calculate the effective radius R of the superconducting loop and this radius was plotted as a function of the magnetic field. The dotted horizontal lines in Figure 43 represent the mean inner Ri and outer radius Ro determined from the scanning electron microscope picture. It is seen that as the magnetic

field intensity is changed from negative to positive values, the effective radius, as defined from the flux quantization condition of the loop, changes from inner to outer radius and vice versa. This is in agreement with the results of [39]. For a superconducting loop at low magnetic field intensities, it is expected that the appropriate effective radius is given by the geometrical mean value of the outer and  inner radius R = Ri Ro [39, 50, 56] (see also Section 6). This is indeed in good agreement with the observed behavior around zero magnetic field (Fig. 43). In the regime of high magnetic field the concept of surface superconductivity becomes important and the giant vortex state will nucleate on the edges of the loop [38, 39, 56]. In this regime two degenerate current carrying situations are possible [48]; hence the giant vortex state can either circulate the loop clockwise or anti-clockwise. Later, Vodolazov et al. [57] studied multiple flux jumps and irreversible behavior of thin Al superconducting rings using ballistic Hall magnetometry. The technique employs small Hall probes microfabricated from a high-mobility 2D electron gas. The rings, having radii R ≃ 1 ;m and width w ranging from 0"1 to 0.3 ;m, were placed directly on top of the microfabricated Hall crosses, which had approximately the same width b of about 2 ;m (see micrograph in Fig. 44 for a ring with an artificial defect, i.e., an identation). These experimental structures were prepared by multistage electron beam lithography with in accuracy of alignment between the stages of better than 100 nm. The rings studied in this work were thermally evaporated and exhibited 2.6 Double Flux Jumps 2.4

Single Flux Jumps

Effective Radius [µm]

2.2 3

∆(µ0H) [1.412 Gauss]

2

1

2.0 1.8 2.6 2.4 2.2

0

2.0 1.8 –100

–1

–50

0

50

100

µ0H [Gauss] –2

–3 –100

–50

0

50

100

µ0H [Gauss]

Figure 42. The magnetic field intensity difference > ;0 H between two successive jumps in magnetization given in units of 1"412 G corresponding to a single flux quantum -0 = h/2e. The plotted jumps are given as a function of magnetic field intensity. The measurement was performed at T = 0"36 K. The positive (negative) flux values correspond to the case in which ;0 H was decreased (increased) during the measurements. Arrows indicate sweep direction. Reprinted with permission from [3], S. Pedersen et al., Phys. Rev. B 64, 104522 (2001). © 2001, American Physical Society.

Figure 43. Effective radius R calculated by using Eq. (58). The data points are the same as those presented in Figure 42. Because the measurements were performed by stepping the magnetic field intensity with a finite step, the effective radius is only measured with a precision of approximately 40 nm. The filled (open) dots correspond to single flux jumps n = 1 (double flux jumps n = 2). The horizontal lines correspond to the outer and inner radius determined from the SEM pictures. The arrows indicate sweep direction. It is seen that as the magnetic field intensity is changed the effective radius changes between inner and outer radius of the loop, a change that depends on sweep direction and magnetic field intensity. The large spread of the data at high magnetic fields corresponds to regions where the amplitude of the oscillations measured by the Hall probe are small. Reprinted with permission from [3], S. Pedersen et al., Phys. Rev. B 64, 104522 (2001). © 2001, American Physical Society.

614

Doubly Connected Mesoscopic Superconductors

2.0

(a)

1.5

–M (arb.units)

1.0 0.5 0.0 –0.5 –1.0 –1.5 –2.0 2.0

(b)

–M (arb.units)

1.5

Figure 44. A micrograph of the superconducting ring placed on top of a Hall bar. An artificial defect (narrowing of the ring cross-section) is intentionally made by electron beam lithography. Reprinted with permission from [57], D. Y. Vodolazov et al., Phys. Rev. B 67, 054506 (2003). © 2003, American Physical Society.

1.0 0.5 0.0 –0.5 –1.0 –1.5 –2.0 –200

11. DYNAMIC TRANSITIONS BETWEEN METASTABLE STATES For systems that exhibit a series of metastable states and that may be brought far from equilibrium, the dynamics will be very important. For such systems the fundamental problem is to determine the final state to which the system will

–100

0

100

200

H (G)

Figure 45. Magnetic field dependence of the magnetization of the ring without (a) and with (b) an artificial defect at T ≃ 0"4 K. Parameters of the rings (width and radii) are the same within experimental accuracy. Reprinted with permission from [57], D. Y. Vodolazov et al., Phys. Rev. B 67, 054506 (2003). © 2003, American Physical Society.

transit (see, e.g., [59]). Vodolazov et al. [58] presented a detailed study of the dynamics in a mesoscopic superconducting ring [58]. This system is a typical example of the above mentioned systems in which a set of metastable states exist. They showed that the final state depends crucially on 3.0

2.5

–M (arb.units)

a superconducting transition at about 1.25 K. The superconducting coherence length was  T = 0 ≃ 0"18 ;m. The latter was calculated from the electron mean free path l ≃ 25 nm of macroscopic Al films evaporated simultaneously with the Al rings. Rings with and without an artificial defect were studied. In Figure 45a and b the full magnetization loop of such a ring with parameters R = 1"0 ± 0"1 ;m and w = 0"25 ± 0"05 ;m is shown for, respectively, a ring without and with an artificial defect at T ≃ 0"4 K. The experimental value for the coherence length obtained from the mean free path is given by  0 ≃ 0"18 ;m. In Figure 46 the low free part of the virgin magnetization curve is presented for a ring without (solid curve) and with (dotted curve) an artificial defect. From the figures one can notice the nonunity of the vorticity jumps for the ring without artificial defects; that is, >L = 3 in the low magnetic field region, >L = 2 in the intermediate H region, and >H = 1 near Hmax . In the ring with approximately the same mean radii and width but containing an intentionally introduced artificial defect, jumps with >L = 1 are mostly observed. To understand the multiflux jumps one has to go beyond theories based on thermodynamical equilibrium states. Not only metastable states have to be considered but also the transitions between them. This necessitates the study of time-dependent phenomena [57, 58] which will be presented in the next section.

2.0

1.5

1.0

0.5

0.0 0

10

20

30

40

50

H (G)

Figure 46. Magnetic field dependence of the virgin magnetization of a ring without (solid curve) and with (dotted curve) an artificial defect. The dotted curve is shifted for clarity by 0.6. Reprinted with permission from [57], D. Y. Vodolazov et al., Phys. Rev. B 67, 054506 (2003). © 2003, American Physical Society.

615

Doubly Connected Mesoscopic Superconductors

> = div Im F ∗  − iA F

(60)

where F = Fei# is the order parameter, the vector potential A is scaled in units -0 / 2 (where -0 is the quantum of magnetic flux), and the coordinates are in units of the coherence length  T . In these units the magnetic field is scaled by Hc2 and the current density, j, by j0 = c-0 /8 2 2 . Time is scaled in units of the Ginzburg–Landau relaxation time QGL = 4'n 2 /c 2 , the electrostatic potential, , is in units of c-0 /8 2 'n ('n is the normal state conductivity), and u is a relaxation constant. In their numerical calculations they used the new variable U = exp −i A dr which guarantees gauge invariance of the vector potential on the grid. They also introduced small white noise P in their system, the size of which is much smaller than the barrier height between the metastable states. The parameter u is considered as an adjustable parameter, which is a measure of the different relaxation times (for example, the relaxation time of the absolute value of the order parameter) in the superconductor. This approach is motivated by the following observations. From an analysis of the general microscopic equations, which are based on the BCS model, the relaxation constant u was determined in two limiting cases: u = 12 for dirty gapless superconductors [60] and u = 5"79 for superconductors with weak depairing [61, 62]. However, this microscopic theory is built on several assumptions, for example, that the electron–electron pairing interaction is of the BCS form, which influences the exact value of u. On the other hand, the stationary and time-dependent Ginzburg–Landau equations, Eqs. (59) and (60), are in some sense more general and their general form does not depend on the specific microscopic model, but the values of the parameters, of course, are determined by the microscopic theories. (For a discussion of this question see also [63, 64] and for a review of different types of time-dependent Ginzburg–Landau equations we refer to [48, 65–67].)

As shown in [68, 69] the transition of

the superconducting ring from a state with vorticity L =  # ds/2 (which in general can be a metastable state) to a state with a different vorticity occurs when the absolute value of the gaugeinvariant momentum p =  # − A reaches the critical value  1 1 pc = √ 1+ (61) 2R2 3 For simplicity, consider that the magnetic field is increased from zero (with initial vorticity of the ring L = 0) to the critical Hc where p becomes equal to pc (for L = 0 one has Hc = 2pc /R). In this case p = −A and Eq. (61) can be written as  R 1 -c /-0 = √ (62) 1+ 2R2 3 For this value of the flux the thermodynamical equilibrium state becomes Leq = Int -c /-0 . For example, if R = 10 one finds Leq = 6 (Fig. 47). The fundamental question one wants to answer is What will be the actual value of the vorticity of the final state? Will it be the thermodynamic equilibrium state or a metastable one? This answer is obtained from a numerical solution of the time-dependent Ginzburg–Landau equations. In the numerical calculation Vodolazov et al. considered two different rings with radius R = 10 and R = 15. These values were chosen such that the final state exhibits several metastable states and, in principle, transitions can occur with jumps in L that are larger than unity. For R = 10, >L may attain values between 1 and 6 and for R = 15 it may range from 1 to 9. Smaller rings have a smaller range of possible >L values. They found that rings larger than R = 15 did not lead to new effects. In their numerical simulations they increased H gradually from zero to a value Hc + >H (they took >H = 0"036Hc for R = 10 and >H = 0"012Hc for R = 15) during a time interval >t, after which the magnetic field was kept constant. The magnetic field >H and time >t range were chosen sufficiently large to speed up the initial time for the nucleation of the phase slip process but still –0.4 L=0

8

7

–0.5 6

–0.6

1 5

F/F0

the ratio between the relaxation time of the absolute value of the order parameter and the relaxation time of the phase of the order parameter. To solve this problem a numerical study of the timedependent Ginzburg–Landau equations was presented in [58]. In their approach the self-field of the ring (which is valid if the width and thickness of the ring are less than , the London penetration length) could be neglected and hence the distribution of the magnetic field and the vector potential are known functions. The time-dependent GL equations in this case are   9F + iF =  − iA 2 F + F 1 − F2 + P (59) u 9t

–0.7 2

4

–0.8 3

–0.9

11.1. One-Dimensional Rings If the width of the ring is much smaller than  and ⊥ = 2 /d (d is the film thickness), it is possible to consider the ring, of radius R, as a one-dimensional object and the strictly one-dimensional Eqs. (59) and (60), which model the dynamics of a ring with small width, have to be solved. The vector potential is equal to A = HR/2, where H is the applied magnetic field.

–1.0 0.0

0.2

0.4

0.6

0.8

1.0

H/HC

Figure 47. Dependence of the free energy of the one-dimensional ring (with R = 10) on the applied magnetic field for different vorticity L. Reprinted with permission from [58], D. Y. Vodolazov and F. M. Peeters, Phys. Rev. B 66, 054537 (2002). © 2002, American Physical Society.

616

Doubly Connected Mesoscopic Superconductors

sufficiently small to model real experimental situations in which H is increased during a time much larger than u · QGL (e.g., for Al QGL T = 0 ≃ 10−11 s). A change of >H and >t, within realistic boundaries, did not have an influence on their final results. Their detailed numerical analysis showed that the vorticity L of the final state of the ring depends on the value of u and is not necessarily equal to Leq . The larger u is, the larger the vorticity after the transition. The final state is reached in the following manner. When the magnetic field increases, the order parameter decreases (inset of Fig. 48a). First, in a single point of the ring a local suppression of the order parameter occurs, which deepens gradually with time during the initial part of the development of the instability. This time scale is taken as the time in which the value of the order parameter decreases from 1 to 0.4 in its minimal point. When the order parameter reaches the value ∼0"4 in its minimal point the process speeds up considerably and the order parameter starts to oscillate in time at this particular point of the ring (Fig. 48a). Also at the same time oscillatory behavior of the superconducting js = Re F ∗ −i − A F 1.0

(a) 0.8

1.0 0.8

L=3

|ψ|

|ψ|

0.6 L=0

0.4

0.6

∆t

0.4

L=5

t0

0.2

0.2

0.0

–400

–200

0

t/τGL

0.0 0.4

(b)

(Fig. 48b) and the normal jn = −  (Fig. 48c) current density is found at the point where the minimal value of the order parameter is found. At other places in the ring such oscillatory behavior is strongly damped (see, e.g., the thin curves in Figure 48). After some time the system evolves to a new stable state, which in the situation of Figure 48 is L = 3 for R = 10 and L = 5 for R = 15. In [58] the dependence of the dynamics of F on u in the minimal point was also investigated. With increasing value of u it was found that (1) the number of phase slips (or equivalently the number of oscillations of the order parameter at its minimal point) increases and hence >L also increases, and (2) the amplitude of those oscillations decreases with increasing u. The above results lead to the conclusion that in a superconducting ring there are two characteristic time scales. First, there is the relaxation time of the absolute value of the order parameter QF . The second time scale is determined by the time between the phase slips (PSs), that is, the period of oscillation of the value of the order parameter in its minimal point. Below it is shown that the latter time is directly proportional to the time of change of the phase of the order parameter Q# and is connected to the relaxation time of the charge imbalance in the system. In Figure 49 the distribution of the absolute value of the order parameter F and the gauge-invariant momentum p are shown near the point where the first PS occurs for a ring with radius R = 10 and u = 3 at different times: just before and after the first PS which occurs at t ≃ 2"7QGL . Before the moment of the phase slip the order parameter decreases while after the PS it increases. To understand this different behavior, Eq. (59) is rewritten separately for the absolute value F and the phase # of the order parameter

0.3

js/j0

u 0.2 0.1

9 2 F 9F = + F 1 − F2 − p2 ) 9t 9s 2 1 9jn 9# = − − " 9t uF2 9s

(63) (64)

0.0

(a) 0.8

(c)

0.3

|ψ|

0.6

jn/j0

0.2

0.4 0.2

0.1

t = 1.4

t = 3.0

t = 2.0

t = 3.5

t = 2.44

t = 4.9

6

–0.1 0

10

20

30

40

t/τGL

p/(Φ0/2πξ)

(b) 0.0

4 2 0 –2

Figure 48. Dependence of the order parameter (a), the superconducting js (b), and the normal jn (c) current density in the point where the minimal (thick curves) and the maximal (thin curves) values of the order parameter are reached as a function of time for u = 3, R = 10 (dotted curves), and R = 15 (solid curves). For both rings we took >t = 100, the time interval over which the magnetic field is increased to its value Hc [inset of (a)]. Reprinted with permission from [58], D. Y. Vodolazov and F. M. Peeters, Phys. Rev. B 66, 054537 (2002). © 2002, American Physical Society.

0.72

0.76

0.80

0.84

0.88

0.92

s/(2πR)

Figure 49. Distribution of the absolute value of the order parameter (a) and gauge-invariant momentum (b) near the phase slip center at different moments of time for a ring of R = 10 and with u = 3. The phase slip occurs at t ≃ 2"7QGL . Reprinted with permission from [58], D. Y. Vodolazov and F. M. Peeters, Phys. Rev. B 66, 054537 (2002). © 2002, American Physical Society.

617

Doubly Connected Mesoscopic Superconductors

Here s is the arc coordinate along the ring, and we used the condition div js + jn = 0. It is obvious from Eq. (63) that if the right-hand side (RHS) of Eq. (63) is negative F decreases in time and if the RHS is positive F will increase in time. Because the second derivative of F is always positive [at least near the phase slip center (Fig. 49a)] the different time dependence of F is governed by the term −p2 F. From Figure 49b it is clear that after the phase slip the value of p is less than before this moment, with practically the same distribution of F. This fact is responsible for an increase of the order parameter just after the moment of the phase slip. But at some moment of time the momentum p can become sufficiently large, making the RHS of Eq. (63) negative and as a consequence F starts to decrease. Based on their numerical calculations they stated that for every value of the order parameter in the minimal point, there exists a critical value of the momentum pcmin such that if the value of the momentum in this point is less than pcmin , the order parameter will increase in time. In the opposite case the order parameter decreases, which leads ultimately to the phase slip process. Therefore, after a phase slip p increases fast enough such that at some moment the condition pmin > pcmin is fulfilled, the order parameter will start to decrease, which leads then to a new phase slip process. Vodolazov et al. [58] concluded that the change of p (or phase of the order parameter) with time and of F with time has a different dependence on u. They found that with increasing u the time relaxation of F becomes larger with respect to the relaxation time for p (e.g., the amplitude of the oscillations in F decreases). Moreover, the relaxation time for p depends not only on u but also on the history of the system: the larger the number of phase slips which have occurred in the system the longer the time becomes between the next two subsequent phase slips. To understand such behavior one has to turn to Eq. (64). Numerical analysis shows that the second term in the RHS of Eq. (64) is only important for some distance close to the PS center. This distance scale is nothing either than the length over which the electric field penetrates the superconductor. This can be checked directly by solving Eqs. (59) and (60) for this situation. But this length is the decay length Q of the charge imbalance in the superconductor (see, e.g., [48, 67]). Numerically it is found that Q varies with u (in the range u = 1 − 100) as Q ∼ u−0"27 , which means that the relaxation time of the charge imbalance is QQ ∼ 2Q ∼ u−0"54 . If we take the derivative 9/9s on both sides of Eq. (64) and integrate it over a distance of Q near the PS center (far from it we have jn ∼ 0) we obtain the Josephson relation (in dimensionless units)

where · means averaging over time between two consecutive phase slips. This result allows one to qualitatively explain the numerical results. Indeed, from Figure 48c it is apparent that jn 0  decreases after each PS and as a result it leads to an increase of Q# . With increasing u, Q decreases and Q# increases. Fitting the data (Fig. 50) leads to the dependencies Q# ∼ u0"21 and Q# ∼ u0"23 for rings with radius R = 10 and R = 15, respectively (here Q# was defined as the time between the first and the second PS). This is close to the expected dependencies that follow from Eq. (66) and from the above dependence of Q u . In addition, Eq. (66) allows us to explain the decrease of Q# with increasing radius of the ring (see inset in Fig. 50). Namely, during the time between two PSs the gauge-invariant momentum decreases as >p ∼ 1/R (because  # increases as ∼2/2R) in the system. Far from the PS center the total current becomes practically equal to js ∼ p. Because div js + jn = 0 in the ring, one finds directly that during this time the normal current density in the point of the PS also decreases as ∼1/R. Taking into account Eq. (66) one can conclude that Q# should vary as ∼1/R (at least for a large radius and for the first PS). The behavior shown in the inset of Figure 50 is very close to such a dependence (it is interesting to note that in contrast to Q# the time QF is practically independent of R). On the basis of their results Vodolazov et al. [58] made the following statement: when the period of oscillation (time of change of the phase of the order parameter) becomes of the order, or larger, than the relaxation time of the absolute value of the order parameter the next PS becomes impossible in the system.

d># = V ∼ jn 0 Q dt

Figure 50. Dependence of the initial nucleation time t0 (curves 1 and 2), the relaxation time of the order parameter QF (which for definiteness we defined as the time of variation of the order parameter from 0"4 until the first PS (curves 3 and 4), and the relaxation time of the phase of the order parameter Q# [we defined it here as the time between the first and second PS (curves 5 and 6)] are shown as function of the parameter u. Dotted (solid) curves are for a ring with radius R = 10 15 . The dashed curve corresponds to the relaxation time of the charge imbalance (as obtained from the expression QQ = 5"792Q ). In the inset, the dependence of Q# on the ring radius is shown for different values of the parameter u. Reprinted with permission from [58], D. Y. Vodolazov and F. M. Peeters, Phys. Rev. B 66, 054537 (2002). © 2002, American Physical Society.

(65)

where ># = # +Q − # −Q is the phase difference over the PS center which leads to the voltage V = −  +Q −  −Q and where jn 0 is the normal current (or electric field in our units) in the point of the phase slip. From Eq. (65) it follows immediately that the relaxation time for the phase of the order parameter near the PS center is Q# ∼

1 Q jn 0 

(66)

1

103

2

t/τGL

τφ/τGL

10

102

24

8

12 6 4

u=6 6

8

101

10

12

14

16

18

R/ξ

3,4 5 6

0

20

60

40

80

100

u

618

external magnetic field the defect leads to a nonuniform distribution of the order parameter, which is not so for the nonuniform ring case.

11.2. Two-Dimensional Rings The results of the previous section were based on a onedimensional model that contains the essential physics of the decay and recovery of the superconducting state in a ring from a metastable state to its final state. However, even when the effect of the self-field can be neglected, the finite width of the ring may still lead to important additional effects (e.g., a finite critical magnetic field). To include the finite width of the ring into our calculation Vodolazov et al. [58] considered the following model. They took a ring of mean radius R = 12, width 3"5, and thickness less than  and . These parameters are close to those of the ring studied experimentally in [3] (see Section X) and they are such that the self-field of the ring can still be neglected. The magnetization curves obtained for such a ring are shown in Figure 51 as a function of the magnetic field and for two values of u. Those results were obtained from a numerical solution of the two-dimensional Ginzburg–Landau equations [Eqs. (59) and (60)]. The magnetic field was changed with steps >H over a time interval that is larger than the initial part of the phase slip process t0 . From Figure 51 it can be noted that the value of the vorticity jumps, >L, depends sensitively on the parameter u. The PSs occur in one particular place along the perimeter of the ring. However, in contrast to the previous onedimensional case, >L depends also on the applied magnetic field. The reason is that for a finite width ring the number of metastable states decreases with increasing magnetic field [70]. It means that the system cannot be moved far from equilibrium with a large superconducting current density (because the order parameter is strongly suppressed by the external field) at high magnetic field. Hence, the value

–M (a.u.)

50

u=3 R = 12ξ

0

w = 3.5ξ

∆L = 1 ∆L = 2

–50

50

–M (a.u.)

Above it was found that the time scale governing the change in the phase does not coincide with the relaxation time of the absolute value of the order parameter. This difference is essentially connected with the presence of an electrostatic potential in the system. To demonstrate this the following numerical experiment was performed. Vodolazov et al. neglected  in Eq. (59) and found that the number of PSs is now independent of u and R, and >L equals unity. Moreover, it turned out that the time scale t0 is an order of magnitude larger. This clearly shows that the electrostatic potential is responsible for the appearance of a second characteristic time which results in the above mentioned effects. In an earlier paper [68] the question of the selection of the metastable state was already discussed for the case of a superconducting ring. However, Tarlie and Elder neglected the effect of the electrostatic potential and found that transitions with >L > 1 can occur only when the magnetic field (called the induced electromoving force (emf) in [68]) increases very quickly. In their case these transitions were connected to the appearance of several nodes in F along the perimeter of the ring and in each node a single PS occurred. Vodolazov et al. [58] reproduced those results and found such transitions also for larger values of >H. However, simple estimates show that to realize such a situation in practice it is necessary to have an extremely large ramp of the magnetic field. For example, for Al mesoscopic samples with  0 = 100 nm and R = 10 the corresponding ramp should be about 103 –104 T/s. With such a ramp the induced normal currents in the ring are so large that heating effects will suppress superconductivity. The transitions between the metastable states in the ring are not only determined by u and R, but also the presence of defects plays a crucial role. Their existence leads to a decrease of >L. This is mainly connected to the fact that with decreasing pc (as a result of the presence of a defect) the value of jn 0  decreases even at the moment of the first PS (because jn 0  cannot be larger than js 0  ≃ pc ) and hence Q# increases. This can also be explained in terms of decreasing degeneracy of the system. For example, if one decreases pc with a factor of two (e.g., by the presence of defects) it leads to a twice smaller value for -c and Leq . But the numerical analysis of Vodolazov et al. showed that the effect of defects is not only restricted to a decrease of pc and Leq . To show this they simulated a defect by including another material with a smaller or zero Tc . This is done by inserting in the RHS of Eq. (59) an additional term @ r F where @ r is zero except inside the defect where @ r =  < 0. The magnetic field was increased up to Hc of the ideal ring case. It was found that the number of PSs was smaller compared with these for a ring without defects. The calculations were done for defects such that pc was decreased by less than a factor of 2 in comparison to the ideal ring case. In contrast, a similar calculation for a ring with nonuniform thickness/width showed that, even for “weak” nonuniformity (which decreases pc by less than 20%), >L was larger than for the ideal ring case and the final vorticity approaches Leq . This remarkable difference between the situation for a defect and the case of a nonuniformity may be traced back to the difference in the distribution of the order parameter: even in the absence of an

Doubly Connected Mesoscopic Superconductors

(a) u=6

0 ∆L = 4

∆L = 3

∆L = 2

–50 –4

∆L = 1

(b) –2

0

2

4

H/(Φ0 /πξR)

Figure 51. Dependence of the magnetization of a finite width ring on the external magnetic field for two values of the parameter u. The results are shown for a sweep up and sweep down of the magnetic field. Reprinted with permission from [58], D. Y. Vodolazov and F. M. Peeters, Phys. Rev. B 66, 054537 (2002). © 2002, American Physical Society.

619

Doubly Connected Mesoscopic Superconductors

Note that now pc decreases with increasing magnetic field. This automatically leads to a decreasing value of the jump in the vorticity >L at high magnetic field, because in [58] it was shown that >Lmax = Nint pc R

–M (arb.units)

20

(a)

10 0 –10 –20 1.0

(b)

|ψ|

0.8 0.6 R = 5.5ξ

0.4

w = 1.5ξ 0.2 0.0

(c)

0.4 0.2

p

of jn 0  will be much smaller in comparison to the one at low magnetic fields, and Q# is larger or comparable with QF even for high values of u and even for the first PS. Thus, the effect of a large magnetic field for a finite width ring is similar in some respects to the effect of defects for a onedimensional ring. The numerical results are in qualitative agreement with the experimental results of Pedersen et al. [3] (see Fig. 41). Unfortunately, no quantitative comparison is possible because of a number of unknowns, for example, the value of pc [it is necessary to have the dependence of M H as obtained starting from zero magnetic field]; the value of  is not accurately known, and hence the ratio R/ can only be estimated. The values of both these parameters have a strong influence on the value of the vorticity jumps >L. For the two-dimensional ring the modified critical momentum can be written as [57]:    1 1 H 2 pc = √ + 1− (67) Hmax 2R2 3

0.0 –0.2

(68)

–0.4

where Nint x is the nearest integer of x. To check the validity of Eq. (67) a numerical simulation of the two-dimensional Ginzburg–Landau equations [Eqs. (59) and (60)], was performed for a ring with R = 5"5 and w = 1"5 (for these parameters the theoretical findings fit the experimental results of Geim et al. [57]). In Figure 52 the magnetization, the order parameter and the gauge-invariant momentum p are shown as function of the applied magnetic field. The magnetic field was cycled up and down from H < −Hmax to H > Hmax . The condition (67) leads to an hysteresis of M H and to a changing value of the jump in the vorticity in accordance with the change in pc . The main difference between the theoretical prediction (67) and the results of the numerical calculations of Vodolazov et al. appears at fields close to Hmax . Apparently it is connected to the fact that for the considered ring the distribution of the order parameter along the width of the ring is appreciably nonuniform at H ≃ Hmax and as a consequence the one-dimensional model breaks down. Finally, Vodolazov et al. also considered the same ring with a defect. The effect of the defect was modelled by introducing in the RHS of Eq. (59) the term −@ s F (s is the arc coordinate) where @ s = −1 inside the defect region with size  and @ s = 0 outside. This leads to the results shown in Figure 53 for M H , F H , and p H . Due to the presence of the defect, pc already differs from Eq. (67) at low magnetic field [pc H = 0 ≃ 0"33 at given “strength” of the defect] and as a result only jumps with >L = 1 are possible in such a ring. In this case the pc and F also depend on the applied magnetic field with practically the same functional dependence on H as Eq. (67). The numerical results of Vodolazov et al. can be compared to the experimental results of Geim et al. [57].

–0.6 –200

–100

0

100

200

H (G)

Figure 52. Magnetic field dependence of the magnetization (a), the order parameter (b), and the gauge invariant momentum  (c) in the middle of the ring. Dotted curve in (b) is the  expression 1 − H /Hmax√ 2 . The dotted curve in (c) is the expression 1 − H − H0 /Hmax 2 / 3, where H0 ≃ 13G is the displacement of the maximum of M H from the H = 0 line. Reprinted with permission from [57], D. Y. Vodolazov et al., Phys. Rev. B 67, 054506 (2003). © 2003, American Physical Society.

Figures 52a and 45a are qualitatively very similar. For example, the theory describes: (1) the hysteresis; (2) the nonunity of the vorticity jumps, that is, >L = 3 in the low magnetic field region, >L = 2 in the intermediate H region, and >L = 1 near Hmax [theoretically (experimentally) one found 6(5), 13(21), and 22(18) jumps with respectively, >L = 3, 2, and 1]; and (3) the nonsymmetric magnetization near ±Hmax for a magnetic field sweep up and down. In the ring with approximately the same mean radii and width but containing an intentionally introduced artificial defect, jumps with >L = 1 are mostly observed (see Fig. 45b). The reason is that an artificial defect considerably decreases the critical value pc (and hence the field Hen ; see dotted curve in Fig. 46). From Figures 52(c) and 53(c) it is clear that the maximum value pcid ≃ 0"54 for a ring without a defect and pcd ≃ 0"35 for a ring with a defect. The ratio pcd /pcid ≃ 0"65 is close to the ratio of the field of the first d id vortex entry Hen /Hen ≃ 0"67 obtained from experiment (see Fig. 46). From Figure 52c it is easy to see that for a ring without a defect at p ≃ 0"35 there are only jumps with >L = 1. But if we slightly increase p then jumps with >L = 2 can appear in the system. So one can conclude that p = 0"35 is

620

Doubly Connected Mesoscopic Superconductors 20

–M (arb.units)

(a) 15

j ds = 2n − f2

10

5

0 1.0

(b) 0.8

|ψ|

0.6 R = 5.5ξ

0.4

w = 1.5ξ

0.2

(c)

0.3 0.2

p

flux - through the ring (which follows from the single valuedness of F; see, [73]),

0.1 0.0 –0.1 –0.2 –0.3 –200

–100

0

100

200

H (G)

Figure 53. Magnetic field dependence of the magnetization (a), the order parameter (b), and the gauge invariant momentum (c) (in the middle of the ring) ofa ring containing one defect. The dotted line in (b) is the expression 1 − H /Hmax 2 . Reprinted with permission from [57], D. Y. Vodolazov et al., Phys. Rev. B 67, 054506 (2003). © 2003, American Physical Society.

close to the border value that separates regimes with jumps in vorticity of >L = 1 and >L = 2. Thermal fluctuations may influence the value of >L and in particular for a pc value close to this border value. This is probably the reason why in the experiment (Fig. 45b) occasional jumps with >L = 2 are observed, which are absent in the simulation (Fig. 53a).

12. PHASE SLIP STATES The nonuniform superconducting state in a ring in which the order parameter vanishes at one point is called the phase slip state. This state is characterized by a jump of the phase by  at the point where the order parameter becomes zero. In uniform rings such a state is a saddle-point state and consequently unstable. However, for nonuniform rings with, for example, variations of geometrical or physical parameters or with attached wires, this state can be stabilized and may be realized experimentally. In recent years [70–72] the existence of a single-connected state in a ring was proposed theoretically. Such a state implies that the relation between the phase # of the order parameter F = fei# , the current density j, and the magnetic

(69)

is no longer valid (the flux is expressed in -0 /2 and the current density in j0 = c-0 /4 2 2 ;  is the London penetration length,  is the coherence length, and -0 is the quantum of magnetic flux). The reason is that the order parameter in such a single-connected state is zero at one point. Moreover, it was claimed that under certain conditions, that is, radius of the ring less than  and the flux through the ring of about n + 1/2 -0 , this state may become metastable in some range of magnetic fields [73]. Together with Vodolazov we revisited this problem [75] and we showed that a state where the order parameter vanishes in one point is still double-connected and that Eq. (69) remains valid for such a state. We propose to call such a state a one-dimensional vortex state (ODV state), because like for an ordinary 2D Abrikosov vortex, there is a jump in the phase of the order parameter of  at the point where F = 0. In contrast to a 2D Abrikosov vortex the ODV is an unstable structure in an uniform ring. However, if some inhomogeneities are present in the ring (defects, nonuniform thickness or nonuniform width of the ring, on attached superconducting wires) the ODV structure can be stabilized and may consequently be realized experimentally. Consider a ring with thickness d   and width w  . In addition, let the radius of the ring R be much larger than w. Under these conditions we can neglect the screening effects and the problem is reduced to a one-dimensional one. The distribution of the current density j and the order parameter F of the system at a temperature not far from the critical temperature Tc is described by the one-dimensional Ginzburg–Landau equation (plus the condition divj = 0) d2 f + f 1 − f 2 − p2 = 0 ds 2 d 2 dj = f p=0 ds ds

(70) (71)

where the gauge-invariant momentum p =  # − A is scaled in units of -0 / 2 , the length of the ring is L = 2R, and the circular coordinate s is in units of the coherence length . In these units, the magnetic field is scaled in units of the second critical field, Hc2 , and the magnetic flux in -0 /2. The coupled Eqs. (70) and (71) have to be solved with the boundary condition F −L/2 = F L/2 . We use the method proposed in [76] (see also [55]). These equations have the first integral   1 df 2 f 2 f4 j2 + − + 2 =E 2 ds 2 4 2f

(72)

From a formal point of view, Eq. (72) is nothing other than the condition of conservation of energy for some “particle” with “coordinate” f and “momentum” j. The role of “time” is played [76] by the circular coordinate s. In

621

Doubly Connected Mesoscopic Superconductors

Figure 54 we show the dependence of the “potential energy” of this system U f =

f4 j2 f2 − + 2 2 4 2f

(73)

for different values of j. Possible solutions of Eqs. (70) and (71) are in the region where confined “trajectories” of our virtualparticle exist. This is possible for currents 0 < j ≤ jc (jc = 4/27j0 is the depairing current density). Using Eq. (72) we can immediately write the solution of Eqs. (70) and (71) for the ring √

2s =



t



(74)

where t s = f 2 s , F C) m is the elliptic integral of the first kind and t0 ≤ t1 ≤ t2 are the solutions of the cubic equation t 3 − 2t 2 + 4Et − 2j 2 = 0

(75)

Using the boundary condition for f we may conclude that for a given current there exist a maximum number of three solutions. There are two uniform solutions [in the points of the minimum E = Emin and the maximum E = Emax of the “potential energy” (73)] and one nonuniform solution with energy Emin < E < Emax , which is defined by the equation    8 t1 − t 0 L= (76) K t2 − t 0 t2 − t 0 where K m is the complete elliptic integral of the first kind. Numerical analysis of Eq. (76) for 0 < j < jc shows that there is a minimal value for the ring length Lmin for which there exists a solution for this equation. When j → jc , Lmin →  and in the opposite limit j → 0 one can show that Lmin → . The latter corresponds to a ring radius

(78) (79)

After inserting these results into Eq. (76) we obtain the “energy.” For L ≫  the “energy” E ≃ 1/4 is practically independent of L [e.g., for L = 8 the difference 1/4 − E L = 8 is less than 0.002]. Also in the limit L −  ≪ 1 we found that the “energy” E is independent of j E L ≃

  1 L− 2 2 /8

(80)

In the limit j ≪ 1 we can also find the dependence of t s near the minimum point of t s t s = 2j 2 + s 2 /2) 2

2

t s = j /2E + 2Es )

L ≫ )

s≪1

L −  ≪ 1)

s≪1

(81) (82)

Using Eqs. (81) and (82) it is easy to show that in the limit j → 0 the gauge-invariant momentum is given by p s = j/t s =

j  + 2m  s j

(83)

where  s is the Dirac delta function and m is an integer. As a result the integral j ds = lim p s ds 2 j→0 j→0 f

= # +5 − # −5 = ># = ±  + 2m (84)

0.5

4

0.4

U

(77)

lim

0.6

0.3

0.2

3

E

2

0.1 1

0.0 0.0

t0 ≃ j 2 /2E) √ t1 ≃ 1 − 1 − 4E √ t2 ≃ 1 + 1 − 4E

dt

t − t0 t1 − t0 t2 − t0     t − t0 t1 − t 0 2 −1 F sin ) = √ t2 − t 0 t1 − t 0 t2 − t 0 t0

R = 1/2 (or /2 in dimensional units). For a radius R > 1/2 metastable states may exist and superconductivity is present for any value of the magnetic flux (at least in our onedimensional model) [73, 77]. In principle, Eqs. (74)–(76) define the nonuniform solution of Eqs. (70) and (71) for a ring. Unfortunately, even using the explicit Kordano expressions for the roots of Eq. (75) the results are rather complicated for arbitrary values of the current j. However a tractable analytical solution is possible if we consider the case for low currents j ≪ 1 for which the roots of Eq. (75) are simplified to

0.2

0.4

0.6

0.8

1.0

1.2

f

Figure 54. Dependence of the “potential energy” U f for different values of the  current j: (1) j = 0"01, (2) j = 0"05, (3) j = 0"15, and (4) j = jc = 4/27. Reprinted with permission from [75], D. Y. Vodolazov et al., Phys. Rev. B 66, 054543 (2002). © 2002, American Physical Society.

for arbitrary ring size. Combining Eq. (84) with Eq. (69) it is easy to show that if - = k + 1/2 #0 (k is an integer) a solution of Eq. (70) can be found which vanishes at one point. In previous studies of this state [71–73], Eq. (83) was not taken into account (only the absolute value of the order parameter was found). Furthermore, in [73] it was claimed that for small rings there is a magnetic field region in which such a state exists and is metastable. But we find that in this region this state even does not exist (except in one point) because Eq. (69) is not fulfilled! The distribution of the order parameter as obtained in [75] may also be applied to a ring in which some part of the ring consists of a normal metal. Then at the boundary between the normal metal and the superconductor the condition F ≃ 0 is fulfilled (if the normal part is longer than )

622

Doubly Connected Mesoscopic Superconductors

and the distribution of the density F s 2 coincides with the one obtained in [73] (but the phase # s will be different for our previous ring geometry). Besides we cannot call such a state single-connected as in [71–73] because the relation (69) is valid even for this case. Therefore, it is better to call this state a ODV state because like the 2D Abrikosov vortex there is a point where F = 0 and the phase of the order parameter exhibits a jump equal to k (as the Abrikosov vortex with orbital momentum 2k). A numerical calculation shows that, as in the case for an ordinary Abrikosov vortex in superconductors where such a state is stable, the phase jump is always equal to ; that is, k = 1. The numerical analysis of the time-dependent Ginzburg– Landau equations showed that this state is unstable for a uniform ring. However for rings with nonuniform width (thickness) or attached wires a nonuniform distribution of the order parameter becomes possible [71, 72, 78, 79] and for small rings with L ∼  this state is realized in practice (besides for the case of a ring, a ODV state can also exist in the Wheatstone bridge [80, 81]). In Figure 55 the distribution of the absolute value and the phase of the order parameter is shown for different values of the magnetic field for a nonuniform ring. We used the model of [71] in which 0.6 .

.

.

.

.

.

.

.

(a)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

0.5

.

.

.

. .

. .

. .

. .

. .

.

.

.

.

.

.

.

. .

. .

.

.

. . .

.

.

.

.

.

.

. .

.

.

.

. . .. . .. .

..

.. .

. .

.. . .. . ..

. .

f

. .

. .

. ..

.. .

.

. . ..

.

..

0.3

.. .. ...

0.8

fmin

0.6 0.4 0.2 0.0 0.0

0.1

.

.

..

1.0

0.2

. . .

.

. .

. .

.

0.4

.

1

. . .

0.2

0.4

.

. . .

. . .

.

. .

. . .

.

. .

. . .

. .

.

.

. . .

.

. . .

. . .

.

.

. .

.

. .

.

. .

.

.

.

.

.

.

.

.

.

.

. .

.

.

0.8

.

.

1.0

.

.

.

Φ/Φ0

.

3

.

. .

.

4

. . . ........ . . .

2.0

.

2

.

.

0.6

. .

.

. . .

.

.

.

(b)

1.0

. . . .

0.5

∆φ/π

1.5

5

φ/π

–1.0 0.0

0.2

0.4

0.6

0.8

.

.

. . .

.

7

. . . .

.

8

.

. . .

.

. .

.

.

.

.

. . .

.

6 .

–0.5

1.0

. .

. . .

.

0.0

.

.

.

. . . . . . . .

.

. . .

.

. . .

.

.

.

1.0

.

.

.

.

.

. .

Φ/Φ0

0.5

. .

.

.

.

.

.

.

.

.

.

.

.

.

. . .

. . . .

.

.

.

.

.

.

.

. . . .

.

.

.

.

.

.

. . .

.

. . .

. . . .

. . . .. . .

. .. .. .

.

. .. .. .

. . .. . . .

.

. .. . .

. . .

.

.

. . . .

. . . .

. . . .

. . .

.

. . .

.

. .

.

.

.

.

.

. . .

. .

.

.

.

.

.

. .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

. . . .

.

.

.

. .

.

. .

.. . .

. .

. .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

. . . .

.

2

.

.

.

.

3

.

4

.

.

.

.

. .

.

.

.

.

.

.

. . .

.

–0.5 0.0

.

.

.

.

. .

.

.

.

.

.

.

1

. .

. .

.

. . . .

. . . .. .

. .

.

.

.

.

. . . .

.

.. ..

. . . .

. . .

.

.

. .

. . .

. . . .

.

.

. . . . . .

.

.

.

.

0.0

.

.

.

.

.

. .

. . . .

. . . . .

. .

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

s/L

Figure 55. Distribution of the absolute value (a) and phase (b) of the order parameter at different values of the magnetic flux: (1) - = 0"45-0 , (2) - = 0"49-0 , (3) - = 0"498-0 , (4) - = 0"5 − 0 -0 , (5) - = 0"5 + 0 -0 , (6) - = 0"502-0 , (7) - = 0"51-0 , and (8) - = 0"55-0 . Note that the order parameter for - = -0 /2 +  is the same as that for - = -0 /2 −  when  < -0 /2. In the insets (a) fmin and (b) the phase difference ># near the point s/L = 3/4 are shown as a function of -. Reprinted with permission from [75], D. Y. Vodolazov et al., Phys. Rev. B 66, 054543 (2002). © 2002, American Physical Society.

the width of the ring was varied as w s = 1 + w0 sin



2s L



(85)

The parameters used are w0 = 0"1 and L = 3"25. It is seen that with increasing magnetic flux the magnitude of the order parameter f decreases in the thinner part of the ring (at s/L = 3/4) and becomes zero when -/-0 = 1/2. In [72] it was found that the ODV state is only possible at - = n + 1/2 -0 . The reason is now clear—only at this value of the magnetic flux were the phase difference near the point where the order parameter is zero be compensated for by the term 2n − - and the current density will be equal to zero. It is interesting to note that the current density is also zero for - = n-0 even in the case of a nonuniform ring, but for values of the flux where the ODV state does not exist. At these values of the magnetic flux the order parameter is uniform along the ring and the term 2n is completely compensated for by the term - in Eq. (69). Above we showed that the ODV state can be realized by varying the geometrical parameters of the ring. But there are two alternative experimental approaches to realize the ODV state in the ring. First, it is possible to include an other phase in the ring. This leads to the appearance of a weak link in the sample and if the radius of the ring is less than some critical value (about ) it also leads to the existence of the ODV state in the ring at - = n + 1/2 -0 . We modeled this situation by introducing an additional term @ s f in Eq. (70), and, as an example, we took @ s = − s and  = 1 (it should be noted that qualitatively the results do not depend on the specific value of ). We will not present the numerical results of the modified Ginzburg–Landau equations, but we found that they are qualitatively similar to the behavior of f and # shown in Figure 56. When approaching the flux - = n + 1/2 -0 the order parameter reaches zero in the defect point and a jump in the phase equal to  occurs. Secondly, such a ODV state should also appear in small rings with attached wire(s). As was shown in [78, 79] an attached wire leads to a nonuniform distribution of the order parameter in the ring. In Figure 56 the dependence of the current j in such a ring on the flux through the ring is shown (see also [79]). As for nonuniform rings, where the parameters chosen such that the ODV state may exist, there is no hysteresis in such a system [72]. In addition, for rings with an attached wire the dependence dj/d- on - has two maxima in the range 0) 1 for small L in contrast to the case for nonuniform rings (see [72]). Small inhomogeneities in rings give us the unique possibility to study states close to the ODV state because there is no sharp transition from the state with finite order parameter to a state with vanishing order parameter at one point (see Fig. 55). The latter is, in some sense, a “frozen” slip phase state. A small variation of the flux through the ring near - = -0 /2 leads to a change of ># by 2 (from − to + with a concomitant change of the current from −0 to +0; see Fig. 55b). As a result an additional phase circulation  #ds = 2 appears in the system, not because the

magnetic flux - is changed, but because the term j/f 2 ds changes from − to + [see Eq. (69)]. When such a jump occurs  # does not change in the ring, except near the point

623

Doubly Connected Mesoscopic Superconductors 0.5 0.0

8

0.4

4

dj/dΦ

0.2

Ro/ξ = 2.0

2 3

4

0.3

–0.2

1

0

Ri/ξ = 1.5

–0.4 0.0

0.2

0.4

0.6

0.8

1.0

0.4 0.3

Φ/Φ0

–0.6

1

–0.8

–M/Hc2

0.0

–8

F/F0

j/j0

0.1

a/ξ = 0.4 d/ξ = 0.001 κ = 2.0

0.5

–4

–0.1

0.2 0.1 0.0

–0.2

–0.1

2 3

–0.3

–0.2 2

0

4

4

–1.0

5

6

8

H/Hc2

–0.4 0

2

4

8

6

H/Hc2 0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Φ/Φ0

Figure 56. The current in the ring as a function of the flux. The different curves are for different circumference of the ring: (1) L = 1, (2) L = 2, (3) L = 3, (4) L = 3"5, and (5) L = 4. For L = 4 the ODV state does not exist in the ring and hysteresis appears. In the inset the dependence of dj/d- on the flux - is shown for rings with lengths L = 1, 2, 3, and 3.5 (curves 1, 2, 3, and 4, respectively). Note that there are two maxima in the range 0) 1 . With increasing length the two maxima merge into one and when one approaches the critical length dj/d- diverge at - = -0 /2. Reprinted with permission from [75], D. Y. Vodolazov et al., Phys. Rev. B 66, 054543 (2002). © 2002, American Physical Society.

where f = 0, and as a result the current density in the system changes continuously. In the usual case transitions of the vorticity from a state with phase circulation 2n to a state with phase circulation 2 n + 1 leads

to a jump in the current everywhere in the ring because  # ds changes by 2 and the order parameter and the current density are finite everywhere in the system. To supplement the above study we also made a numerical study of a nonuniform ring of finite width and thickness by implementing our previous finite difference solution of the coupled nonlinear Ginzburg–Landau equations [52]. As an example we took the following parameters: outer radius of the ring Ro = 2, radius of the hole Ri = 1"5, displacement of the hole from the center a = 0"4, ring thickness d = 0"001, and Ginzburg–Landau parameter  = 2. In Figure 57 the dependence of the Gibbs free energy F and the magnetization M = −9F /9H of this system on the magnetic field are shown. As in the case of our one-dimensional ring these dependencies are reversible, and there are magnetic fields at which the magnetization is equal to zero (at these points the free energy has a local maximum). We found that the distribution of the order parameter and the phase in the ring (see inset of Fig. 58) is similar to analogous distributions for the above one-dimensional ring (Fig. 55) at low magnetic fields. However, a difference with the onedimensional system is that the phase circulation increase of 2 [between points (2) and (3) in Fig. 58] does not occur at the magnetic field value where the free energy has a local maximum (and zero magnetization). With decreasing width of the ring this point shifts toward the local maximum in the free energy. It is interesting to note that for the system corresponding to Figure 57 the vortex enters through the thinnest part

Figure 57. The Gibbs free energy and the magnetization (right lower inset) of the asymmetric ring (upper inset) as function of the applied magnetic field. Reprinted with permission from [75], D. Y. Vodolazov et al., Phys. Rev. B 66, 054543 (2002). © 2002, American Physical Society.

of the ring for the first three maxima in the free energy while for the highest magnetic field maxima (i.e., H /Hc2 ≃ 4"6) the vortex enters through the thickest part of the ring (see Fig. 59). Because the width of the thicker part of the ring (=0"9) is considerable larger than that of the thinner part (=0"1) and is of order , the ODV for H /Hc2 ≃ 4"6 becomes the usual two-dimensional vortex. As a result the order parameter is equal to zero only in one point along the radial coordinate and the circulation of the phase of the order parameter now is also a function of this coordinate (see Fig. 59). When the vortex enters/exits the ring at low magnetic fields there is also a slow dependence of the order parameter on the radial coordinate along the thinnest part of the ring. Therefore, we can conclude that in a ring with finite width the one-dimensional vortex is transformed into the usual two-dimensional vortex.

–0.8366

(4)

–0.8368

2.0

(3)

L=0→1

(a)

1.5

–0.8370

φ/π

0.2

(2)

–0.8372

1.0

(1) (3)

0.5

(2) (4)

0.0

–0.8374

–0.5 1.0

–0.8376

0.8

|Ψ|

0.1

F/F0

–0.5 0.0

–0.8378

–0.8382

0.6 0.4

–0.8380

0.2

(b)

0.0 0.0

(1)

0.2

0.4

0.6

0.8

s/L

–0.8384 0.650

0.655

0.660

0.665

0.670

0.675

H/Hc2

Figure 58. The Gibbs free energy F H near the first maximum. In the inset the phase and the order parameter distribution at different values of the applied magnetic field [indicated by the squares on the F H curve] are shown. The phase was calculated along the outer perimeter of the ring. The open circle on F H indicates the position at which the vorticity increases from 0 to 1. Reprinted with permission from [75], D. Y. Vodolazov et al., Phys. Rev. B 66, 054543 (2002). © 2002, American Physical Society.

624

Doubly Connected Mesoscopic Superconductors –0.0490

–0.2 Louter = 3 → 4

(a)

–0.0495

(b)

(c)

(d)

–0.3

Linner = 3 → 4

–0.4 a/ξ = 0.0 –0.5

a/ξ = 0.1 a/ξ = 0.2

–0.6

–0.0505

(a)

F/F0

F/F0

–0.0500

(b)

a/ξ = 0.3 –0.7

a/ξ = 0.4

–0.0510 –0.8 –0.0515

–0.9 L=1

–1.0 –0.0520 4.59

(c) 4.60

4.61

4.62

(d) 4.63

L=0

4.64

H/Hc2

Figure 59. The same as Figure 58 but now for the last maximum in Figure 57. In the inset a contour plot of the order parameter is shown for different values of the magnetic field which are given by the squares in the main figure. Louter is the vorticity as calculated along the outer perimeter of the ring and Linner is the vorticity calculated along the perimeter of the hole. Reprinted with permission from [75], D. Y. Vodolazov et al., Phys. Rev. B 66, 054543 (2002). © 2002, American Physical Society.

The process of the appearance of a stable vortex in a ring with finite width is similar to the creation of a phase slip in wires of finite width in the presence of a transport current. In the latter case the distribution of the gauge-invariant momentum is not exactly uniform over the wire width. The maximum value of p is obtained at the edges and as a result the order parameter vanishes first in these points. When a phase slip is created the distribution of the order parameter is not uniform over the width. With decreasing wire width this nonuniformity decreases, but it will be uniform, strictly speaking, only in the limit w → 0. With increasing wire width the phase slip transforms to the ordinary process of vortex/antivortex pair nucleating at the edges, penetrating deep into the superconductor and annihilating. In nonuniform rings the distribution of p over the width is nonuniform and in contrast to wires with transport current, it is nonsymmetric with respect to the ring width. Suppression of the order parameter first occurs only on the outer side (when increasing the magnetic field) or on the inner side (when the magnetic field is decreasing). As in the case of an one-dimensional ring we may call this state a stable (“frozen”) phase slip state if the width of the ring, where the vortex penetrates, is less than . The ODV state of a nonuniform ring is very similar to the saddle-point state as found in [52] for a uniform ring of finite width and in [8] for the case of a disk. Note that for an uniform ring with zero width the nonuniform solution [Eq. (75)] corresponds to a saddle point of F F . In Figure 60 the gradual decrease of the hysteresis with increasing displacement a of the hole from the center of the ring is shown. It is seen that with increasing a the region where metastable states exist decreases and ultimately vanishes for some critical displacement ac . The saddle point states (only shown for a = 0, by the thin full curve) approach the stable phase slip state when a = ac . In [39] nonuniform superconducting rings of finite width were also investigated, but with a larger ring radius (see also

–1.1 0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

H0 /Hc2

Figure 60. The Gibbs free energy F H at low magnetic fields for different displacements a of the hole from the center of the ring. The thin solid curve (for a = 0) corresponds to the saddle point state. Reprinted with permission from [75], D. Y. Vodolazov et al., Phys. Rev. B 66, 054543 (2002). © 2002, American Physical Society.

Section 8). It was found that for low magnetic fields F H is irreversible but for high magnetic fields the dependence F H was reversible. In high fields the transition from a state with vorticity one to another state occurred through the same scenario as discussed above. A similar state was also considered in [82, 83] in the framework of the linearized Ginzburg–Landau equations for a nonuniform ring of finite width. In such a system the vortex may be stable in some magnetic field region. In [83] analytical results for the magnetic field ranges where the vortex enter through the thinner part of the ring and through the thicker part were obtained. Our calculations generalize this result to lower temperatures (i.e., where the nonlinear Ginzburg–Landau equations have to be used) and with inclusion of the nonzero demagnetization factor of the ring. We found dependence of the Gibbs free energy (and magnetization) of this sample on the applied magnetic field. Our results also allowed us to compare the results of the one-dimensional model with the full two-dimensional one. The ODV (or stable phase slip) state may be observed by magnetic experiments. Magnetic susceptibility is proportional [55] to dj/d- and the ODV state exhibits some characteristic peculiarities as was shown in the inset of Figure 56. Magnetization M is proportional to the current j and hence M H is reversible, for samples where the ODV state exists (see Fig. 56 and inset of Fig. 58). Furthermore M = 0 at - ≃ n + 1/2 -0 . Alternatively, because at - = n + 1/2 -0 there is a point in the ring where the order parameter is equal to zero, this state may be found by transport measurements. For example, in [84] the dependence of the resistance of a hollow cylinder with radius of order  was studied at temperature T < Tc far from Tc . At - = -0 /2 the resistance R exhibited a maximum but the value of R was a factor of 3 less than the normal state resistance Rn . If the cylinder has a nonuniform thickness the ODV state can be realized as a stable state when - = -0 /2, and it will lead to a resistive (but superconducting) state even for very small currents and naturally the resistance of such a state will be less than Rn as was found by Liu et al. [84].

Doubly Connected Mesoscopic Superconductors

GLOSSARY Giant vortex state State in which all the vortices are combined in the center into one big vortex corresponding with one minimum in the Cooper pair density. Mesoscopic superconductors Superconductors with sizes comparable to the coherence length and the penetration depth. Multivortex state State in which all the vortices are separated from each other. Phase slip center The point in a quasi-one dimensional (with cross-section less than ) superconductor where a jump by 2 in the phase occurs when the order parameter becomes equal zero. The appearance of a phase slip center in superconducting wires with transport current leads to a finite resistivity of the sample.

ACKNOWLEDGMENTS This work was supported by the Flemish Science Foundation (FWO-Vl), the “Onderzoeksraad van de Universiteit Antwerpen” (GOA), the “Interuniversity Poles of Attraction Program—Belgian State, Prime Minister’s Office—Federal Office for Scientific, Technical and Cultural Affairs” (IUAPVI), and the European ESF-network on Vortex Matter. One of us (B.J.B.) is a postdoctoral researcher with the FWO-Vl. We thank the late V. Schweigert, S. Yampolskii, A. Geim, and D. Vodolazov for collaboration on the different subjects discussed in this chapter. Stimulating discussions with V. Moshchalkov are gratefully acknowledged.

REFERENCES 1. A. K. Geim, I. V. Grigorieva, S. V. Dubonos, J. G. S. Lok, J. C. Maan, A. E. Filippov, and F. M. Peeters, Nature (London) 390, 256 (1997). 2. O. Buisson, P. Gandit, R. Rammal, Y. Y. Wang, and B. Pannetier, Phys. Lett. A 150, 36 (1990). 3. S. Pedersen, G. R. Kofod, J. C. Hollingbery, C. B. Sørensen, and P. E. Lindelof, Phys. Rev. B 64, 104522 (2001). 4. R. Benoist and W. Zwerger, Z. Phys. B 103, 377 (1997). 5. J. J. Palacios, Physica B 256–258, 610 (1998). 6. J. J. Palacios, Phys. Rev. B 58, R5948 (1998). 7. J. J. Palacios, Phys. Rev. Lett. 84, 1796 (2000). 8. V. A. Schweigert and F. M. Peeters, Phys. Rev. Lett. 83, 2409 (1999). 9. S. V. Yampolskii and F. M. Peeters, Phys. Rev. B 62, 9663 (2000). 10. E. Akkermans, D. M. Gangardt, and K. Mallick, Phys. Rev. B 62, 12427 (2000). 11. E. Akkermans and K. Mallick, Physica C 332, 250 (2000). 12. E. Akkermans and K. Mallick, J. Phys. A 32, 7133 (1999). 13. P. Singha Deo, V. A. Schweigert, F. M. Peeters, and A. K. Geim, Phys. Rev. Lett. 79, 4653 (1997). 14. P. Singha Deo, F. M. Peeters, and V. A. Schweigert, Supperlattice Microstruct. 25, 1195 (1999). 15. V. A. Schweigert and F. M. Peeters, Phys. Rev. B 57, 13817 (1998). 16. V. A. Schweigert, F. M. Peeters, and P. Singha Deo, Phys. Rev. Lett. 81, 2783 (1998). 17. V. A. Schweigert and F. M. Peeters, Physica C 332, 426 (2000). 18. V. A. Schweigert and F. M. Peeters, Phys. Rev. B 60, 3084 (1999). 19. V. V. Moshchalkov, L. Gielen, C. Strunk, R. Jonckheere, X. Qiu, C. Van Haesendonck, and Y. Bruynseraede, Nature (London) 373, 319 (1995).

625 20. V. Bruyndoncx, J. G. Rodrigo, T. Puig, L. Van Look, V. V. Moshchalkov, and R. Jonckheere, Phys. Rev. B 60, 4285 (1999). 21. H. T. Jadallah, J. Rubinstein, and P. Sternberg, Phys. Rev. Lett. 82, 2935 (1999). 22. L. F. Chibotaru, A. Ceulemans, V. Bruyndoncx and V. V. Moshchalkov, Nature (London) 408, 833 (2000). 23. L. F. Chibotaru, A. Ceulemans, V. Bruyndoncx, and V. V. Moshchalkov, Phys. Rev. Lett. 86, 1323 (2001). 24. J. Bonˇca and V. V. Kabanov, Phys. Rev. B 65, 012509 (2002). 25. B. J. Baelus and F. M. Peeters, Phys. Rev. B 65, 104515 (2002). 26. V. R. Misko, V. M. Fomin, J. T. Devreese, and V. V. Moshchalkov, Physica C 369, 361 (2002). 27. V. R. Misko, V. M. Fomin, J. T. Devreese, and V. V. Moshchalkov, Phys. Rev. Lett. 90, 147003 (2003). 28. V. L. Ginzburg and L. D. Landau, Zh. Eksp Teor. Fiz. 20, 1064 (1950). 29. V. V. Schmidt, P. Müller, and A. V. Ustinov, “The Physics of Superconductors: Introduction to Fundamentals and Applications.” Springer-Verlag, Berlin/Heidelberg, 1997. 30. D. Saint-James, G. Sarma, and E. J. Thomas, “Type II Superconductivity.” Pergamon Press, Oxford, 1969. 31. P. G. de Gennes, “Superconducting of Metals and Alloys.” Benjamin, New York, 1966. 32. A. C. Rose-Innes and E. H. Rhoderick, “Introduction to Superconductivity.” Pergamon Press, Oxford, 1969. 33. A. A. Abrikosov, Sov. Phys. JETP 5, 1175 (1957). 34. G. F. Zharkov, V. G. Zharkov, and A. Yu. Zvetkov, Phys. Rev. B 61, 12293 (2000). 35. W. A. Little and R. D. Parks, Phys. Rev. Lett. 9, 9 (1962). 36. R. D. Parks and W. A. Little, Phys. Rev. 133, A97 (1964). 37. F. London, “Superfluids.” Wiley, New York, 1950. 38. V. Bruyndoncx, L. Van Look, M. Verschuere, and V. V. Moshchalkov, Phys. Rev. B 60, 10468 (1999). 39. B. J. Baelus, F. M. Peeters, and V. A. Schweigert, Phys. Rev. B 61, 9734 (2000). 40. V. V. Moshchalkov, M. Dhalle, and Y. Bruynseraede, Physica C 207, 307 (1993). 41. A. K. Geim, S. V. Dubonos, J. G. S. Lok, M. Henini, and J. C. Maan, Nature (London) 396, 144 (1998). 42. R. P. Groff and R. D. Parks, Phys. Rev. 176, 567 (1968). 43. H. J. Fink, Phys. Rev. 177, 732 (1969). 44. F. M. Peeters and X. Q. Li, Appl. Phys. Lett. 72, 572 (1998). 45. V. M. Fomin, J. T. Devreese, and V. V. Moshchalkov, Europhys. Lett. 42, 553 (1998). 46. V. M. Fomin, J. T. Devreese, and V. V. Moshchalkov, Europhys. Lett. 46, 118 (1999). 47. A. P. van Gelden, Phys. Rev. Lett. 20, 1435 (1968). 48. M. Tinkham, “Introduction to Superconductivity.” McGraw Hill, New York, 1975. 49. G. F. Zharkov, in “Superconductivity, Superdiamagnetism, Superfluidity” (V. L. Ginzburg, ed.), p 126. MIR, Moscow, 1987. 50. R. M. Arutunyan and G. F. Zharkov, J. Low Temp. Phys. 52, 409 (1983). 51. R. Kato, Y. Enomoto, and S. Maekawa, Phys. Rev. B 47, 8016 (1993). 52. B. J. Baelus, F. M. Schweigert, and V. A. Schweigert, Phys. Rev. B 63, 144517 (2001). 53. P. Singha Deo, V. A. Schweigert, and F. M. Peeters, Phys. Rev. B 59, 6039 (1999). 54. C. P. Bean and J. D. Livingston, Phys. Rev. Lett. 12, 14 (1964). 55. X. Zhang and J. C. Price, Phys. Rev. B 55, 3128 (1997). 56. F. M. Peeters, V. A. Schweigert, B. J. Baelus, and P. S. Deo, Physica C 332, 255 (2000). 57. D. Y. Vodolazov, F. M. Peeters, S. V. Dubonos, and A. K. Geim, Phys. Rev. B, 67, 054506. 58. D. Y. Vodolazov and F. M. Peeters, Phys. Rev. B 66, 054537 (2002). 59. M. C. Cross and P. C. Hohenberg, Rev. Mod. Phys. 65, 851 (1993).

626 60. L. P. Gor’kov and G. M. Eliashberg, Zh. Eks. Teor. Fiz. 54, 612 (1968) [Sov. Phys. JETP 27, 328 (1968)]. 61. A. Schmid, Phys. Kondens. Mater. 5, 302 (1996). 62. L. Kramer and R. J. Watts-Tobin, Phys. Rev. Lett. 40, 1041 (1978). 63. T. J. Rieger, D. J. Scalapino, and J. E. Mercereau, Phys. Rev. B 6, 1734 (1972). 64. A. T. Dorsey, Phys. Rev. B 46, 8376 (1992). 65. B. I. Ivlev and N. B. Kopnin, Usp. Fiz. Nauk 142, 435 (1984) [Sov. Phys. Usp. 27, 206 (1984)]. 66. A. L. de Lozanne and M. R. Beasley, “Nonequilibrium Superconductivity” (D. N. Langenberg and A. I. Larkin, Eds.), p. 111. North-Holland, Amsterdam, 1986. 67. A. M. Kadin and A. M. Goldman, “Nonequilibrium Superconductivity” (D. N. Langenberg and A. I. Larkin, Eds.), p. 253. NorthHolland, Amsterdam, 1986. 68. M. B. Tarlie and K. R. Elder, Phys. Rev. Lett. 81, 18 (1998). 69. D. Y. Vodolazov and F. M. Peeters, Physica C 400/3-4, 165 (2003). 70. J. Berger, Phys. Rev. B 67, 014531 (2003). 71. J. Berger and J. Rubinstein, Phys. Rev. Lett. 75, 320 (1995).

Doubly Connected Mesoscopic Superconductors 72. J. Berger and J. Rubinstein, Phys. Rev. Lett. 56, 5124 (1997). 73. E. M. Horane, J. J. Castro, G. C. Buscaglia, and A. Lopez, Phys. Rev. B 53, 9296 (1995). 74. M. Tinkham, Phys. Rev. 129, 2413 (1963). 75. D. Y. Vodolazov, B. J. Baelus, and F. M. Peeters, Phys. Rev. B 66, 054543 (2002). 76. J. S. Langer and V. Ambegaokar, Phys. Rev. 164, 498 (1967). 77. A. Bezryadin, A. Buzdin, and B. Pannetier, Phys. Rev. B 51, 3718 (1995). 78. H. J. Fink and V. Grunfeld, Phys. Rev. B 33, 6088 (1986). 79. H. J. Fink and V. Grunfeld, Phys. Rev. B 31, 600 (1985). 80. H. J. Fink, Phys. Rev. B 45, 4799 (1992); 48, 3579 (1993). 81. C. Ammann, P. Erdös, and S. B. Haley, Phys. Rev. B 51, 11739 (1995). 82. J. Berger and J. Rubinstein, Phys. Rev. B 59, 8896 (1999). 83. J. Berger, “Connectivity and Superconductivity” (J. Berger and J. Rubinstein, Eds.), p. 138. Springer-Verlag, New York, 2000. 84. Y. Liu, Yu. Zadorozhny, M. M. Rosario, B. Y. Rock, P. T. Carrigan, and H. Wang, Science (Washington, DC) 294, 2332 (2001).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Drug Nanocrystals of Poorly Soluble Drugs Rainer H. Müller Pharmasol GmbH, Berlin, Germany

Aslihan Akkar Free University of Berlin, Berlin, Germany

CONTENTS 1. Introduction 2. General Features of Drug Nanocrystals 3. “Bottom-Up” Technologies for Drug Nanocrystal Production (Hydrosols, NanoMorphTM ) 4. “Top-Down” Technologies 5. Combination of “Bottom-Up” and “Top-Down” Technology (Nanoedge® ) 6. Perspectives of the Drug Nanocrystal Technology Glossary References

1. INTRODUCTION Poor solubility of a drug is in most cases associated with poor bioavailability, at least bioavailability problems. About 10% of the present drugs are poorly soluble, about 40% of the drugs in the pipeline possess a poor solubility, and even 60% of drugs coming directly from synthesis have a solubility below 0.1 mg/mL [1]. The problem is even more pronounced than in the past because more and more new drugs are poorly soluble in aqueous media and simultaneously in organic media. This excludes previous formulation approaches such as dissolution of the drugs in solvent mixtures (e.g., ethanol/water) or dissolving them in oils or the lipophilic phase of microemulsions. In the case of too low oral bioavailability of poorly soluble drugs, in most cases, parenteral administration is not an alternative. Injection volumes are either too high because of the very low solubility, or the drugs cannot be dissolved at all. The injection of crystalline microparticulate drug dispersions would

ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

even worsen the situation compared to oral administration because the body fluid volume at the injection site is very limited. From this, new smart technological formulations for improved delivery are required to overcome this problem of most newly developed drugs, and to make them available to the patient. Improved formulations are even necessary at a much earlier state in the drug discovery process. A pharmacological screening for activity and side effects is not possible without having reached a sufficient availability in the blood. According to the biopharmaceutical classification, drugs with poor solubility fall either in class II or class IV. Class II drugs show poor solubility, but good permeability, which means that their bioavailability problems can be overcome when improving the solubility. Class IV drugs are characterized by simultaneously poor solubility and low permeability. In this case, improving the solubility will not or will little improve the bioavailability. Therefore, the attempts to find improved formulations are mainly targeting class II drugs. The “specific” formulation approaches are characterized in that they are working only with a very limited number of specific molecules which “fit.” For example, molecules need to fit in size to be complexed by cyclodextrins or for efficient solubilization by mixed micelles. The molecules need to have the right structure to sufficienty interact with the surfactants forming the micelles. This limited applicability is clearly proven by the very low number of products on the market being based on such formulation principles. Much more attractive are the “generalized” formulation approaches; a very simple and effective one is micronization. Any drug can be micronized independent of its structure. In the past, this worked for quite a number of drugs; however, it is not effective anymore for the new drugs exhibiting even lower saturation solubilities. Consequently, the next step after “micronization” was “nanonization,” which means that, instead, producing ultrafine drug microparticles (mean diameter approximately 2–10m), one produces

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (627–638)

628

Drug Nanocrystals of Poorly Soluble Drugs

superfine drug nanocrystals (approximately 100 nm to just below 1000 nm). To be more precise, a drug showing poor solubility means that it has a low saturation solubility which is typically correlated with a low dissolution velocity dc/dt. The terms “solubility” and “saturation solubility” are used synonymously. In case a drug—despite having a low saturation solubility—dissolves very fast and has a good permeability, this should not cause any bioavailability problems. However, this is normally not the case with poorly soluble drugs because, according to the law of Noyes–Whitney, the dissolution velocity dc/dt is directly correlated to the saturation solubility cs . The principle of micronization is to increase the surface area A to which dc/dt is also directly correlated. Nanonization is taking the increase in surface area A one dimension further than micronization. There are also some additional features of drug nanocrystals improving the dissolution velocity (increase in dissolution pressure, reduction of diffusional distance h), discussed below, which further improve the enhancement of bioavailability. There is a limited number of technologies around to produce drug nanocrystals; basically, one can differentiate between “bottom-up” techniques and “top-down” techniques. All of them are patent protected, which means that it is a “closed market” and dominated by only a few companies. The different technologies are briefly discussed and compared, and advantages and disadvantages highlighted. The main focus is on drug nanoparticles produced by milling or by high-pressure homogenization using preferentially piston-gap homogenizers. Special emphasis is given to the large-scale production, including regulatory aspects— important criteria for entering the market with a product.

2. GENERAL FEATURES OF DRUG NANOCRYSTALS The law of Noyes–Whitney describes the dissolution velocity [2]: DA dm =

c − cx dt h S

(1)

D is the diffusion coefficient, A is the surface area, cs is the saturation solubility, cx is the bulk concentration, and h is the so-called “diffusional distance” over which the concentration gradient occurs (note: division of the equation by the volume v leads to dc/dt). It is obvious that an increase in A increases the dissolution velocity. To illustrate this, the surface area was calculated for a 100 mg dose of drug assuming that the drug powder is a finely milled powder of 50 m particles, micronized powder of 5 m particles, and nanonized powder of 200 nm particles (assumed drug density 2.0 g/cm3 : • 50 m particles: 603 × 10−3 m2 • 5 m particles: 0.0599 m2 • 200 nm particles: 1.50 m2 . The Kelvin equation [3] describes the vapor pressure above very fine liquid droplets in a gas phase, for example, during the spray drying process. It describes the vapor pressure as

a function of the droplet radius, which means as a function of the droplet curvature: ln

2Mr Pr = P rRT

(2)

where Pr is the dissolution pressure of a particle with the radius r, P is the dissolution pressure of an infinitely large particle,  is the surface tension, R is the gas constant, T is the absolute temperature, r is the radius of the particle, Mr is the molecular weight, and  is the density of the particle. The basic information is that the vapor pressure above a flat surface and a slightly curved surface are identical; going to a strongly curved surface (i.e., typically particles below 1–2 m) leads to an increase in vapor pressure (Fig. 1). The situation at the liquid–gas interface can directly be transferred to the situation at the solid–liquid interface, which means that the vapor pressure of the liquid corresponds to the dissolution pressure of the solid (drug). It is known that different liquids have different vapor pressures; for example, comparing ether, water, and oil (Fig. 2), the same compound-specific differences are, of course, valid for solid materials. Consequently, different drugs have different dissolution pressures at identical particle size. Based on the Kelvin equation, the relative increase in vapor pressure as a function of decreasing particle size was calculated [4]; as model compounds, three liquids were chosen having distinct differences in vapor pressure. Oleic acid has a relatively low vapor pressure (boiling point: 350  C) [5], water has a relatively high vapor pressure (boiling point: 100  C) [6], and ether, as a highly volatile liquid, possesses the highest vapor pressure (boiling point: 35  C) [7]. The vapor pressure was normalized, which means that only the relative increase was calculated when reducing the size leading to characteristic functions (Fig. 2). Water and ether represent compounds of good and very good solubility; the relative increase takes place below the known threshold size of approximately 1.0 m, but is not very pronounced. Oleic acid represents the poorly soluble drug (low vapor pressure = low dissolution pressure). The first slight increase is already observed below approximately 2.0 m, being very distinct below the well-known 1.0 m and highest below 200 nm (Fig. 2, upper curve). It is noteworthy that the increase in relative pressure going from 200 nm Dissolution pressure (∆p ~ 1/r)

Substance layer (flat surface)

CsN > CsM

Large particle (slight curvature)

Nanosuspension (strong curvature)

Figure 1. Vapor pressure above differently curved surfaces of liquids occurring at the liquid–gas interface, corresponding to the dissolution pressure at the interface solid–liquid, as a function of the curvature (drug, CsM /CsN —saturation solubility of microparticles and nanoparticles, respectively). Modified with permission from [10], R. H. Müller et al., “Emulsions and Nanosuspensions for the Formulation of Poorly Soluble Drugs,” 1998. © 1998, Medpharm Scientific Publishers.

629

Relative vapor pressure at 293 K

rate refers to the respective surface unit. Based on the surface unit, the intrinsic dissolution rates for both sugars are identical. Based on the outline above regarding the increases in dissolution pressure, the increase in saturation solubility and the reduction in diffusional distance h, the intrinsic dissolution rate increase when one moves in the range below approximately 1 m. Therefore, the nanonization not only increases the measured dissolution rate due to enlargement of the surface A, but additionally the intrinsic dissolution rate as well [9, 10]. It should be noted that the described effects are valid for any identically sized drug nanocrystal independent of its method of production. It does not matter if the drug nanocrystal has been produced, for example, by precipitation, pearl milling, or high-pressure homogenization. For a crystal of a given size (and identical physical state, crystallinity), the same physical effects occur. Therefore, no production method can claim that its nanocrystal has a special feature because it is produced in a certain way. However, differences in identically sized drug nanocrystals can occur when they show differences in their crystalline character, which means the polymorphic form or the fraction of an amorphous drug. It is well documented in the literature that amorphous drugs have a higher saturation solubility than crystalline forms. Therefore, in theory, it is beneficial to have an amorphous drug nanocrystal. However, there are two major challenges for these systems:

Relative increase in vapor pressure [%]

Drug Nanocrystals of Poorly Soluble Drugs

9

oleic acid

8 7 6 5 4 3 2 1

ether water

0 –1 0,1

1

10

Droplet size [µm]

Figure 2. Compound-specific differences in the vapor pressure (dissolution pressure) of different liquids (different drugs) as a function of droplet (particle) size.

down to 100 nm is similarly large as going from 1.0 m to 200 nm. That means, to obtain extremely fast dissolving particles in the blood, a size of 100 nm (or even below) would be ideal. According to the textbooks, the saturation solubility is a compound-specific constant depending only on the temperature. This is valid for the powders of daily life. However, as outlined above, it changes below a particle size of approximately 1–2 m. Saturation solubility is an equilibrium of molecules dissolving and molecules recrystallizing; the increase in dissolution pressure shifts the equilibrium, thus resulting in a higher saturation solubility. Therefore, drug nanoparticles are characterized by an increase in saturation solubility cs . According to Noyes–Whitney, the increase in cs —in addition to A—further increases the dissolution velocity. The final saturation solubility achieved is, of course, compound-specific based on the differences in compoundspecific dissolution pressures (cf. Fig. 2). The dissolution velocity is reversly proportional to the diffusional distance hH , which means that reducing hH leads to a further increase in dissolution velocity. According to the Prandtl equation, the diffusional distance hH decreases for very small particles [8]: hH = k L1/2 /V 1/2

(3)

In this equation, hH is the hydrodynamic boundary layer thickness, k is a constant, L is the length of the surface in the direction of the flow, and V is the relative velocity of the flowing liquid against a flat surface. However, due to some parameters which cannot be specified as actual figures, the decrease in hH can hardly be calculated. Therefore, publications dealing with hH focus rather on direct measurement of the increase in dissolution velocity to have an idea of the quantity of this effect [8]. To clarify the definitions, we must differentiate between dissolution rate (velocity) and the so-called “intrinsic dissolution rate.” The dissolution rate just quantifies the observed dissolution velocity of the powder; therefore, for example, rock candy sugar has a slower dissolution rate than the same amount of sugar in a crystalline form. However, when expressing it as an intrinsic dissolution rate, the intrinsic dissolution rates are identical because the intrinsic dissolution

1. to provide a reproducible and functioning technology to produce 100% amorphous particles, and 2. to guarantee its unchanged amorphous character during the shelf life of the product. There is the general feeling that things are much easier when one immediately begins with a 100% crystalline nanoparticle. The question of shelf-life stability is therefore the major concern with regard to these technologies for amorphous particles—despite the very attractive feature of a further increase in dissolution rate. From the viewpoint of physics, the increase in saturation solubility of amorphous products can be explained by the Ostwald–Freundlich equation. The Ostwald–Freundlich equation is derived from the Kelvin equation, and describes the saturation solubility as a function of the particle size [11]:   S2 4 1 1 RT (4) ln = −

1 − n M S1 ! d2 d1 The parameters are as follows:  n R M S1 S2

= = = = = = = " = d1 = d2 =

dissociation grade number of ions gas constant molecular weight saturation concentration of population 1 saturation concentration of population 2 interfacial energy density diameter of population 1 diameter of population 2.

In a simplified way, amorphous products can be considered as powders with a very small (indefinite) particle size, thus

630 exhibiting a higher saturation solubility compared to crystalline particles. As an aside, to complete the theoretical considerations, the Ostwald–Freundlich equation also contains the interfacial tension  (solid/liquid interface). Thus, it explains and describes the different saturation solubilities obtained with polymorphic forms. The polymorphic forms have different energy contents; consequently, they are different in the interfacial tension at the solid/liquid interface, thus resulting in different saturation solubilities. It should also be considered that energy input to create drug nanocrystals can theoretically lead to a change in polymorphic forms. In tabletting technology, it is described that high compression pressures can induce a transform of a lower energetic form to a higher energetic form (e.g., partial transition from I to II) [12]. Therefore, when producing drug nanocrystals, intensive characterization of the crystalline character after production and—in the case of totally or partially amorphous nanoparticles or nanocrystals of higher polymorphs—the monitoring of the crystalline status during the shelf life of the product are important. Again, the most convenient approach is to create drug nanocrystals that are completely crystalline, and preferentially having the polymorph I. It should be noted that about 50% of the drugs exhibit polymorphism; therefore, this characterization is of great importance.

3. “BOTTOM-UP” TECHNOLOGIES FOR DRUG NANOCRYSTAL PRODUCTION (HYDROSOLS, NanoMorphTM ) Bottom-up technology means that one starts from the molecular level, and goes via molecular association to the formation of a solid particle. That means that we are discussing classical precipitation techniques by reducing the solvent quality, for example, by pouring the solvent into a nonsolvent or changing the temperature or a combination of both. Precipitation is a classical technique in pharmaceutical chemistry and technology. The Latin terminology is via humida paratum (v.h.p.) [13], which means being prepared via a liquid process (solutions are made to obtain a fine powder (precipitate) dispersed in a “wet” environment). Wellknown old examples of the pharmacopea are the mercury ointments (e.g., Unguentum Hydrargyri cinereum, German Pharmacopea no.6, 1926). Because it is a long-known process in pharmacy, a clear outline is necessary to document the innovative height of patent applications based on this process. The soundness of a patent regarding such a technology (of course, the soundness of patents in general), and especially in case they are challenged, is a very important factor for companies when deciding to place a formulation development on such a technology. The so-called “hydrosols” were developed by Sucker and co-workers [14–17] at the Sandoz Company (nowadays Novartis). The drug is dissolved in a solvent, and the solvent is added to a nonsolvent to initiate the fast precipitation of a finely dispersed product. To achieve this, it is necessary to pass the so-called Ostwald–Mier area very quickly, which means reducing the solvent quality very quickly [18]. This is achieved by adding the solvent to a nonsolvent; doing it the

Drug Nanocrystals of Poorly Soluble Drugs

other way around would lead to larger crystals. The formed crystals need to be stabilized by surfactants or polymers to avoid growth of the nanocrystals to microcrystals. In general, it is recommended to lyophilize the product to preserve the nanometer character of the particles [16]. The technology is to some extent complex (e.g., the preservation of particle size), and excludes all molecules which are poorly soluble in aqueous and organic media. To our knowledge, no products based on this technology are marketed by Novartis. An additional problem is, for example, solvent residues; removal makes the process more costly. The particles generated by precipitation, as, for example, by Sucker, are in most cases crystalline in nature; in contrast to this, the company Knoll (nowadays owned by Abbott) creates amorphous particles by a precipitation technique [19]. The product is NanoMorphTM . As outlined above, an additional positive feature is a further increase in the dissolution velocity due to the amorphous character of the product. The precipitation in the amorphous form is achieved by an aqueous polymer solution (company brochure; www.soliqs.com). However, for a commercial product, it is necessary to preserve the amorphous character during the shelf life to avoid changes in bioavailability caused by a reduction in the dissolution velocity due to the transfer of the amorphous drug to a crystalline drug. A precipitation technology was also described by Müller and co-workers (Kiel University, Germany), dissolving the drug in a water-miscible solvent, for example, isopropanol or acetone [20, 21]. The particles are stabilized in the nanometer range by selecting stabilizers possessing optimal affinity to the particle surface, and stabilizing the precipitated particles in the nanometer range. Examples of stabilizers are agar, calcium caseinate, dextran 200, gelatin A, hydroxyethylcellulose, hydroxyethylstarch, hydroxypropylcellulose, polyvinylpyrrolidone, and polyvinylalcohol [22]. The optimum stabilizer might differ from one drug to the next because each drug nanocrystal shows differences in surface properties, such as the presence of functional groups on the surface, surface hydrophobicity, and surface wetability. To transform the particles into a dry product, spray drying is described [23, 24]. The advantage of the precipitation techniques is that relatively simple equipment can be used. For example, the solvent can be poured into the nonsolvent with a constant velocity in the presence of a high-speed stirrer. Very elegant approaches are the use of static mixers or micromixers [25, 26] which simulate the precipitation conditions in a small volume (i.e., simulating lab scale conditions), but being simultaneously a continuous process. In the case of micromixers, scaling up can be performed in a simple way just by so-called “numbering up,” which means arranging many micromixers in parallel [27, 28]. The equipment is relatively simple, and in the case of large conventional static mixers, of relatively low cost (this is not necessarily valid for the micromixers). However, there are some major general disadvantages of the precipitation techniques. 1. The drug needs to be soluble in at least one solvent (thus excluding all new drugs that are simultaneously poorly soluble in aqueous and in organic media). 2. The solvent needs to be miscible with at least one nonsolvent.

631

Drug Nanocrystals of Poorly Soluble Drugs

3. Solvent residues need to be removed, thus increasing production costs. 4. It is a little bit tricky to preserve the particle character (i.e., size, especially the amorphous fraction). In general, it is recommended that a second consecutive process be performed for particle preservation, that is, spray drying or lyophilization.

4. “TOP-DOWN” TECHNOLOGIES There are basically two “top-down” techniques, which means starting from a large powder and performing a size diminution: 1. pearl ball milling [13] and 2. high pressure homogenization. The milling technique was developed and patented by the Nanosystems Company, which was acquired by the élan Company a few years ago. There are two techniques based on high-pressure homogenization in water: the microfluidizer technology developed by the RTP Company [29], and the DissoCubes® technology developed by Mülller et al. [30], nowadays both owned by the company SkyePharma PLC (www.skyepharma.com). Recently, SkyePharma PLC acquired RTP Pharma, which means that the high-pressure homogenization in water is on hand. The third technique is homogenization in media, other than 100% water, which means the Nanopure technology, which is currently owned by Pharmasol GmbH (www.pharmasol-berlin.com).

4.1. Pearl Milling (NanocrystalsTM ) Liversidge and co-workers developed the milling process to yield the so-called (NanocrystalsTM [31]. The mills used by the élan Company are basically containers with small pearls, beads, or balls which can have different sizes, a typical size being 1–2 mm (Fig. 3). The drug powder is dispersed in a surfactant solution, and the obtained suspension is poured into the mill. In the milling process, the pearls are moved, either by using a stirrer or by moving the mill container itself. The drug particles are disintegrated between the moving pearls. This process is accompanied by the erosion of milling material from the pearls, a phenomenon well known for these mills and documented in the literature [32].

stirrer pearls

Figure 3. Principle of pearl mill type; pearls are moved by the stirrer.

Typical milling materials are glass or zirconium oxide. Buchmann et al. reported on the generation of glass particles with a size of a few micrometers during the milling process. The accepted contamination in a product originating from the production equipment used is normally limited in the very low parts-per-million range (e.g., 10 ppm). Depending on the desired small size, the hardness and the crystallinity of the material, and milling forces present, the milling procedure can last from hours or days up to one week. Of course, the contamination increases with the milling time and hardness of the drug material. Some years ago, the NanoSystems Company replaced zirconium oxide by hard polystyrene beads as the milling material. The milling technique clearly has some advantages: 1. simple technology 2. low-cost process regarding the milling itself 3. large-scale production possible to some extent (batch process). Large-scale production is stated only to “some extent.” This should be explained by a model calculation. A pearl mill of 1000 L volume containing milling pearls in hexagonal packaging has 76% pearls (760 L) and 24% void volume available for the drug suspension (240 L). The solid concentration in the suspension is limited; it should not be too pasty because such a highly viscous system is no longer processable. Assuming a suspension with 20% solid would mean that 20% of the 240 L suspension is drug (48 L). This is equivalent to 96 kg of drug (assumed drug density: 2.0 g/mL). Depending on the density of the milling material used (glass: approximately 2–3 g/mL, zirconium oxide: 5.9 g/mL, and hard polystyrene: 1.05 g/mL), the weight of the milling material and suspension will be in the range of 1–2 tons plus the weight of the mill itself. The batch cannot be increased by just multiplying the dimensions of pearl mills [33]. As disadvantages, we can see: 1. potential erosion from the milling material leading to product contamination 2. duration of the process not being very production friendly 3. potential growth of germs in the water phase when milling for a long time 4. time and costs associated with the separation procedure of the milling material from the drug nanoparticle suspension, especially when producing parenteral sterile products. The contamination issue is controversially discussed between the believers in the technology and colleagues who are more skeptical. The company SkyePharma PLC presented data wherein they found 70 ppm contamination in a drug suspension when milling with zirconium oxide pearls [34]. Recently, the élan Company made an oral presentation about intravenous drug nanocrystal suspensions, giving a figure of 70 ppm solid contamination (polystyrene) in the produced drug nanosuspension [35]. The solid contamination was determined by dissolving all of the other parts of the particle suspension, and subsequent filtering of the solution through a 0.2 m filter to collect nondissolved contamination (i.e., eroded polystyrene). It should be kept in mind that the

632 concentration of the contamination increases when removing the water phase of a nanocrystal suspension to obtain a dry product. Of course, the contamination is “diluted” when incorporating the drug nanocrystals into the matrix of a tablet or of pellets. Regarding the figure given by élan, it was not specified how hard the drug was and how long the milling time was, two factors affecting the extent of contamination. Irrespective of the discussion of the contamination issue, élan was successful in getting the first oral product approved, Rapamune® . According to the Rapamune® information package, the tablet is the “dry” alternative to an orally administered solution of Rapamune® . Compared to the solution, the oral bioavailability of the nanocrystal tablet is even slightly higher (approximately 27% higher relative to the oral solution). The bioavailabilities of the nanocrystal tablet and solution are in a similar range; production of the nanocrystal tablet is therefore mainly seen for reasons of convenience, which means easier administration, and less as a final approach to reach sufficiently high bioavailability of a poorly soluble drug. Independent of this, from our point of view, Rapamune® can be seen as a very important milestone because it demonstrates that the drug nanoparticles/nanocrystals are a suitable technology for the pharmaceutical market. It is an important proof of principle for the drug nanocrystal itself—independent of which technology was used to produce it.

4.2. Homogenization in Water (DissoCubes® ) Homogenization in water is performed either using the Microfluidizer® as a high-pressure homogenizer (RTP) [36] or, alternatively, piston-gap homogenizers (SkyePharma PLC) [30]. The Microfluidizer® (Fig. 4) realizes the jet stream principle, which means that two streams of liquid/ suspension hit each other on the front, leading to particle disintegration [26]; different types of interaction chambers are available (“Y” type and “Z” type) (Fig. 5, upper and lower part, respectively). The smallest particle size obtained depends on the power density achieved in the disintegration volume. The power density is defined as power/volume unit (watt/m3 . A measure for the power is the homogenization pressure. Homogenization pressure in Microfluidizer® machines are, for example, 40,000 psi/2760 bar, and for the lab scale mode 120B Microfluidizer® , 15,000 psi/1035 bar. Maximum pressures achievable with piston-gap homogenizers are, for example, 1600 bar with the LAB 40 and Gaulin 5.5, and 2000 bar with the Rannie 118 from APV (APV, Unna, Germany). The Stansted Company can produce homogenizers up to 3500 bar and above [37]. From this, in principle, with given material, the microfluidizer 120B cannot produce such a small particle size as the other lab scale homogenizers with their higher power densities. The examples in the Mikrofluidizer patent [36] described the use of 50 or even up to 100 homogenization cycles to produce drug nanoparticles. This leads to high production times, and is not friendly for the production process. From our investigations, we observed disintegration to real drug nanoparticles only with “soft” drugs. For hard crystalline drugs, the mean diameter stayed in the very low micrometer range. According to the theoretical considerations above

Drug Nanocrystals of Poorly Soluble Drugs

Pre-dispersion

High pressure pump cooler

1st Interaction chamber

2nd Interaction chamber

Figure 4. Basic principle of Microfluidizer® with the first interaction chamber plus second interaction chamber.

(cf. Section 2), this does not utilize all of the benefits of the real drug nanoparticles. In addition, it might be easier to obtain similar particle sizes with other techniques. In the case of highly efficient jet milling, laser diffractometery particle diameters of 2–3 m (diameter 50%) can be obtained. Regarding potential product contamination, no published data were available to us when writing this chapter. Clear disadvantages can be summarized [38]: 1. a relatively high number of required homogenization cycles (up to 100), leaving the production technology user unfriendly 2. a potentially larger content of microparticle diameters above 1 m, not fully providing real drug nanoparticles. Piston-gap homogenizers are based on a different homogenization principle. Liquid/suspension is streaming through a narrow gap with a very high streaming velocity. A typical arrangement for this technology is to place the dispersion (emulsion or suspension) in a cylinder with a piston; a pressure is applied to the piston, pressing the suspension through a gap at the end of the cylinder (Fig. 6). In the Micron LAB 40, the cylinder has a diameter of 3.0 cm, and the narrow gap just has a width of only 25 m; this narrowing by a factor of 120,000 leads to extremely high flow velocities up to 200–300 m/s [39]. For some homogenizer types, velocities up to 1000 m/s are given. According to the patent of DissoCubes® , the reason for disintegration of the particles is the cavitation occurring in this gap. According to the Hagen–Poiseulle equation, the flow rate in the system

633

Drug Nanocrystals of Poorly Soluble Drugs (upper)

In the gap, the dynamic pressure drastically increases (i.e., also streaming velocity), and simultaneously, the static pressure decreases and falls below the vapor pressure of the water at room temperature (Fig. 7). This leads to the boiling of the water at room temperature, the formation of gas bubbles which implode when the liquid leaves the gap being under the normal pressure conditions again (=cavitation). The formation of gas bubbles and their implosion cause shock waves, leading to the destruction of the droplets in the case of an emulsion and the diminution of particles in the case of a suspension. Nowadays, bubble formation and not the implosion is considered as the major reason for diminution. There are advantages of the piston-gap homogenization technique: 1. the effective particle diminution 2. production lines can be qualified and validated 3. production lines already exist in industry, for example, production of parenterals, and so on 4. no contamination, for example, iron content is below 1 ppm [41] 5. off-shelf equipment 6. simple process 7. low-cost process/low-cost equipment.

(lower)

Figure 5. The “Y” interaction chamber (upper) and the “Z” interaction chamber (lower) of the Microfluidizer® . Adapted with permission from [26], R. H. Müller and B. H. L. Böhm, “Dispersion Techniques for Laboratory and Industrial Scale Processing,” 2001. © 2001, Wissenschaftliche Verlagsgesellschaft mbH.

of different diameters is equal at each diameter. According to the Bernoulli equation, the sum of static and dynamic pressure is constant at each diameter of the system [40]: Ps +

1 V 2 = K 2

(5)

where PS is the static pressure, 1/2V 2 is the dynamic pressure (Pdyn ), and K is a constant. Product collection

Gap in form of a ring

To summarize: homogenization with piston-gap homogenizers has good potential to be used as a production technique because it fulfills the key industrial features such as large-scale production and various regulatory aspects [42–44]. The definition includes the term “homogenization” reduction of particle or droplet size and simultaneous narrowing of the size distribution, that is, making the particle population not only smaller, but simultaneously more homogeneous in size [45, 46]. Figure 8 shows an EM graph of a drug compound produced by high-pressure homogenization; the crystals are relatively homogeneous in size and of rather cuboid shape. The homogeneity in size is an essential prerequisite for the physical long-term stability, which means avoiding Ostwald ripening and crystal growth. Ostwald ripening is caused by different saturation solubilities in the vicinity of differently sized particles (cf. aspects of dissolution pressure and saturation solubility of differently sized particles outlined above). Figure 9 shows the situation in the case where very differently sized particles are present. Each particle is surrounded by a layer of saturated solution. Due to the smaller size, the saturation solubility in the vicinity of the small particle (CsN ) is higher then the saturation solubility in the vicinity of the microparticle (CsM ),

Ps

Ps

Pre-emulsion 3 cm

Ps Pdyn

Pdyn

Pdyn

Gap in homogenizer Figure 6. Principle arrangement in piston-gap homogenizer of LAB type (note: some parts are out of proportion (piston, gap) for reasons of clarity).

Figure 7. Change of static pressure (Ps and dynamic pressure (Pdyn as a function of tube diameter according to Bernoulli.

634

Drug Nanocrystals of Poorly Soluble Drugs

little increase in solubility with an increase in temperature, which means that temperature fluctuations during storage will not lead to relevant increases or decreases with fluctuating temperature. There is a general experience; when differently sized particles are partially dissolved by an increase in temperature, recrystallization occurring during a subsequent temperature decrease will preferentially take place on the large crystals, which means that, during temperature cycling processes, small particles dissolve and the larger ones grow.

4.3. Homogenization in Water–Liquid Mixtures and Nonaqueous Media (Nanopure® ) 1 µm Figure 8. EM graph of the model drug RMBB 98. Modified with permission from [61], B. H. L. Böhm, Ph. D. Dissertation, 1999. © 1999, Free University of Berlin.

which means that there is a concentration gradient between the higher concentrated solution around the small particle to the vicinity of the larger particle. Based on the gradient, drug diffuses from the surrounding of the small particle to the surrounding of the large particle. In this case, in the surrounding of the small particle, the solution is no longer saturated, which means that more drug dissolves, and the size of the small particles reduces. Due to drug diffusion in the vicinity of the large particle, a supersaturated solution results, which leads to precipitation of the drug to the particle surface and crystal growth of the large particle. To avoid this, it is necessary to have a uniform particle size, which is achieved by the homogenization process. In the last ten years of our drug nanocrystal research, we never observed crystal growth in a nanosuspension produced by homogenization. In case larger particles occurred during storage, they were only aggregates of small drug nanocrystals due to a suboptimal choice of surfactant or surfactant mixture. The absence of Ostwald ripening is an amazing effect when considering the high disperitivity and energy of the system. Ostwald ripening is promoted in case a drug has a higher solubility, and also shows pronounced solubility increases/decreases as a function of temperature. Nanocrystals are made from poorly soluble drugs; therefore, the saturation solubility is a priori low, and in general, they show CsN > CsM

CsM

CsN

Molecules diffuse across gradient further dissolution

The DissoCubes® patent describes homogenization in pure water as a dispersion medium; the rationale behind that is— as explained in detail in the patent [30]—the cavitation is seen as being responsible for the diminution of the particles. To obtain cavitation in the dispersion medium, one needs to have a high vapor pressure of the dispersion fluid at room temperature. This is fulfilled with water, but not with other liquids such as oils or liquid polyethylene glycol (PEG) 600. In addition, homogenization at higher temperatures (e.g., 80  C) should be much more efficient than at room temperature because the vapor pressure of water is much higher at 80  C, thus leading to more extensive cavitation. The basics of the Nanopure® technology [47] is that homogenization was performed against these teachings, leading to a comparable or even improved product. Under the Nanopure® technology, not only drugs, but also polymers can be processed, leading to a product [48–56] with a mean diameter in the nanometer range (nanoparticles) or a size of a few micrometers (both particle ranges are covered). At the beginning of the Nanopure® development, drug suspensions in nonaqueous media were prepared, for example, drug nanocrystals of cyclosporine (Table 1) and of RMKP 22, the latter drug also being described in the DissoCubes® patent [30]. Also for RMKP 22, a finely dispersed product was achieved (Table 2). Based on the state of the art that homogenization with 100% water leads to a finely dispersed product, after homogenization with 0% water (e.g., liquid PEG), the range with a water content between 0 and 100% water was investigated. The drug RMKP22 was dispersed in dispersion media with increasing water content, which means starting with glycerol (0% water) and going up in water concentration stepwise. An increase in the water content should promote cavitation, and thus lead to more efficient diminution. The result was that no product-relevant differences were found (Table 3 [47]). As pointed out above, the Nanopure® technology covers drugs and polymers to be transferred to nanoparticles or Table 1. PCS characterization data of 1 and 5% cyclosporine drug nanocrystal dispersion (dispersion medium: 1% Tween 80 in propylenglycol).

super saturated solution recrystallization

Figure 9. Principle of Ostwald ripening and crystals growth (for explanation, see text).

Cyclosporine concentration (%) 1 5

PCS diameter (nm)

Polydispersity index

203 182

0132 0131

635

Drug Nanocrystals of Poorly Soluble Drugs 10

RMKP22 concentration (%) 1 3

D 50% (m)

D 95% (m)

094 124

274 378

microparticles with a size of a few micrometers. The teaching in the patent literature [57] was that disintegration of polymeric material is more efficient: 1. when higher temperatures above the glass transition temperature of the polymers are employed, for example, 80  C, and 2. that the use of plasticizers promotes the diminution of polymers. The rationale behind these two points was that soft polymers can be easier disrupted. Polymers are getting softer above their glass transition temperature or when admixing a plasticizer. In the Nanopure® technology, just the opposite was done, which means reducing the temperature below room temperature. Examples were prepared at different temperatures, showing that there was no product-relevant difference in the obtained particle sizes of polymer particles (Fig. 10). Based on these results, high-pressure homogenization at and below the freezing point of water was performed, for example, charcoal homogenized at −20  C. From the data, one can conclude that the temperature has no obvious effect. It rather appears that, by using the low temperature, polymeric material is less elastic and more brittle; nonelastic material appears easier to be broken than a flexible, more elastic one such as a softened polymer. What are the advantages of suspensions produced by the Nanopure® technology? Nonaqueous dispersion media can be used to produce oral formulations, for example, drug nanocrystals dispersed in liquid PEG 600 or miglyol/MCT oil can be directly filled into soft gelatin capsules or sealed hard gelatine capsules. Nanosuspensions in mixtures of water with water-miscible liquids (e.g., ethanol, isopropanol) can be used in the granulation process for tablet production; the solvent mixture evaporates much better than water. Drugs that are susceptible to hydrolysis in water can be prepared as nonaqueous suspensions, for example, glycerol or propyleneglycol. Prior to intravenous injection, the glycerol drug nanosuspensions can be diluted with water for injection to yield an isotonic suspension. Alternatively, isotonic water– glycerol mixtures can be directly homogenized. In the first patent [30], cavitation was considered as the dominating factor for particle diminution; therefore, homogenization was Table 3. Drug RMKP22 homogenized in water-free medium (glycerol) and media of increasing water content (glycerol–water). % water 0% (=100% glycerol) 30% 50% 100% (=pure water)

D 50% (m)

D 95% (m)

128 133 119 094

385 358 309 277

particle size [µm]

Table 2. LD diameters D 50% and D 95% of RMKP 22 homogenized in aqueous surfactant solution at 1500 bar (ten cycles).

8 D 50%

6

D 90% 4

D 95%

2 0

20°C

40°C

60°C

85°C

Figure 10. Particle size characterization data of the polymer ethylcellulose (laser diffractometery data, diameters 50, 90, and 95%) produced by high-pressure homogenization at different temperatures. Adapted with permission from [47], R. H. Müller et al., PCT Application PCT/EP00/06535, 2000. © 2000,

only performed in pure water, and glycerol was added afterwards to avoid an impairment of particle diminution efficiency. Another option of the Nanopure® technology is the production of aqueous polymer particle dispersions with a diameter of a few micrometers. This could be an alternative to polymer dispersion (e.g., Surelease, ethylcellulose) for the coating of tablets which are now produced by using solvents. Aqueous shellack dispersions produced have already been described [58].

5. COMBINATION OF “BOTTOM-UP” AND “TOP-DOWN” TECHNOLOGY (Nanoedge® ) In 2001, the Baxter Company presented their Nanoedge® technology at the annual meeting of the American Association of Pharmaceutical Scientists (AAPS) in Denver, Co. Nanoedge® is an aqueous suspension of drug nanoparticles suitable for intravenous injection. The particles are prepared basically by a high-pressure homogenization process, which means either by microfluidization (RTP technology), homogenization in water using piston-gap homogenizers [30], or alternatively, by a new production process developed by Baxter [59]. This process is called the “microprecipitation method,” and is a combination of precipitation followed by a process with energy input (e.g., homogenization). According to [59], the process includes three steps: 1. dissolving the organic compound in a water-miscible first solvent to form a solution 2. mixing this solution with the second solvent to obtain a “presuspension” 3. adding energy to this “presuspension” to form particles having an average effective particle size of 400 nm– 2.0 m. The method can be applied to all drugs that are poorly water soluble, that is, possess a solubility which is greater in the water-miscible first solvent compared to the second solvent which comprises an aqueous phase. Examples for the first solvent are, for example, methanol, ethanol, glycerol, and acetone. Adding the solvent to the nonsolvent is a classical precipitation process. A second energy-addition step follows, which leads to a particle suspension being crystalline in

636 nature and having an improved physical stability compared to a precipitated product. According to the patent, the process can be separated into three general categories; in addition there are two basic preparation methods, A and B. The steps of the process are identical for each of the categories, the categories themselves are distinguished based on the physical properties of the drug after precipitation (prior to the energy-addition step) and after the energy-addition step. Category I: The drug particles in the presuspension are in an amorphous form, semicrystalline form, or a supercooled melt. Category II: The drug in the presuspension is in a crystalline form. Category III: The organic compound is in a crystalline form that has many imperfections, which means, it is friable. In each of the three categories, after the energy-addition step, the drug nanocrystals are in a crystalline form, which means a physically stable low-energy form with few imperfections. The energy-addition step is more effective, and leads to smaller particles than the drug crystals in the presuspension which are, for example, in an amorphous form (category I) or a crystalline form with a high friability (category III). The physical properties of the drug particles are characterized by DSC or X-ray. Which category of particles is formed depends on many factors, for example, the chemical nature and melting point of the drug, precipitation rate, surfactants present, and precipitation temperature. By adding energy, the physical stability of the suspension is improved, that is, by altering the lattice structure of the crystal to a lowenergy form. In addition, the added energy seems to have an effect on the coating (surfactant layer) of the particle, leading to a rearrangement and improved stabilizing effect. The latter is concluded when comparing the physical stability of precipitated suspensions with suspensions being exposed to an additional energy-addition step, both dispersions having a similar effective particle size between 200 nm and 2.0 m. In method A, the drug is dissolved in the first solvent. A second aqueous solution is prepared containing one or more surfactants; if necessary, the pH is adjusted to yield optimum precipitation conditions. Then the first solution is added to the second solution, the addition rate is chosen depending on the batch size and precipitation kinetics for the organic compound, and the solutions are under constant agitation. If desired, the resulting particles can undergo an annealing step to convert category I particles (amorphous, semicrystalline, and supercooled) into a more stable solid state. In the next step, energy is added, for example, by subjecting the presuspension to sonication, homogenization (piston-gap), microfluidization, or other suitable methods providing cavitation, shear, or impact forces. Homogenizers used are, for example, the Avestin apparatus EmulsiFlex-C 160 to promote a desired phase change into a more perfect crystal; the sample might be subjected to higher temperatures from about 30 to 100  C during the energy-addition step (annealing step). That means that annealing is basically performed by adding energy (homogenization, etc.), and additionally by variation of temperature.

Drug Nanocrystals of Poorly Soluble Drugs

In method B—different from method A—the surfactant or surfactant mixture is added to the first solution. In the second step, the first solution is added to the second solution. This leads to a formation of more uniform, finer particles compared to having the surfactant in the second solution [60]. Drug examples processed are itraconazole and carbamazepine. A disadvantage of the microprecipitation method by Baxter is that additional process steps are required compared to direct homogenization of jet-milled drug powder (e.g., Nanopure® technology). The drug needs to be dissolved, precipitated, and then the solvent needs to be removed again, including analysis for solvent residues. The advantages of the Baxter method are that the problem of the precipitation technique—continuing crystal growth after precipitation—has been overcome because the particles are only an intermediate product to undergo subsequent diminution/annealing. In addition, by optimal choice of precipitation conditions, more friable particles (semicrystalline, amorphous) can be produced which allow more effective diminution in the subsequent high-pressure homogenization step. This is of special importance when the available jet-milled drug is highly crystalline and simultaneously possesses a very hard crystal structure (high Mohs degree). In such cases, the direct homogenization of the jet-milled powder can require a higher number of homogenization cycles, that is, longer production times. In the case of very hard crystals, the maximum homogenization pressure applicable during production (e.g., 2000 bar) might potentially lead to a maximum dispersitivity being rather close to 1 m, and not in the preferred nanometer range. At present, Baxter has various homogenization technologies, which means, depending on the properties of the drug and the desired target, the optimal process method can be chosen to minimize production time and costs.

6. PERSPECTIVES OF THE DRUG NANOCRYSTAL TECHNOLOGY As outlined above, there is a definite need to find a smart formulation to overcome the bioavailability problems of poorly soluble drugs. There will be a huge market for such a formulation when looking at the percentages of the drugs being poorly soluble. Preferentially, the formulation solution should also cover drugs which are poorly soluble in aqueous media and simultaneously in nonaqueous, for example, organic, media. Nanonization, that is, production of drug nanocrystals by bottom-down technologies, definitely appears as the most promising approach. The striking reasons for this is that it is a general technology which can be applied to any drug independent of its solubility properties; in addition, the production process is simple in the case of bottom-down techniques. The area of pearl/ball milling is extensively covered by élan/Nanosystems. The technology has a potential to be used for oral products. Contamination issues might be less of a problem when the milling times are short in the case of less hard crystalline drugs or when contamination is removed by a purification process. The feasibility has been proven by the tablet Rapamune® being on the market. The technology

637

Drug Nanocrystals of Poorly Soluble Drugs

appears more critical for parenteral products, for example, IV, because higher regulatory requirements need to be met. An alternative production technique, homogenization in water, has been covered by the previously competing companies RTP (Microfluidizer® and SkyePharma PLC (pistongap homogenizer). This situation was clarified in the first half of 2002 by SkyePharma PLC finally taking over RTP. By the takeover, competition was reduced, leaving only two in this market of drug nanocrystals. The Baxter Company is also using homogenization in water for its product Nanoedge® . Nanopure® technology is the new third way to produce drug nanocrystals in this densely covered market. Nanopure® allows producing any product as produced by SkyePharma PLC just by simply admixing water with a nonaqueous medium, for example, a mixture of 10% ethanol and 90% water. The ethanol is not critical at all for oral products because, generally, the suspensions are spray dried. In the case of using glycerol–water mixtures (975 + 25), directly isotonic suspension for intravenous injection is obtained. In addition, it has some interesting features when homogenizing in water-free media. Drug nanocrystal suspensions in miglyol/MCT oil or PEG 600 can directly be filled into soft gelatine capsules. The drug nanocrystal suspension can also be produced in melted PEG 6000; after solidification of the PEG 6000, the drug nanocrystals are fixed, and the powder can be coarsely milled and filled into hard gelatine capsules. The coverage of low temperature in water-free media is of interest for temperature-sensitive drugs or molecules being susceptible to hydrolysis or/and high temperature. To summarize, Nanopure® provides the possibility to make at least an identical product or even an improved product. Drug nanocrystal suspensions which are not long term stable as aqueous suspensions can be formulated as glycerol or oily suspensions. The Nanopure® technologies indeed provide an alternative choice for companies with formulation problems, in addition to the other technologies on the market. There is a definite need for a novel formulation of poorly soluble drugs. In the laboratories, the drug nanocrystals— independent of which top-down technology is employed for their production—proved to be such a formulation; many convincing data are available. The final criterion for a technology is its feasibility, proven by the appearance of products on the market. Polymeric nanoparticles have been investigated for almost 30 years; despite this huge input, to our knowledge, no product is on the market (only polymeric microparticles). In contrast to this, the first nanoparticle/ nanocrystal product Rapamune® has recently been launched. Due to the huge number of drug candidates in the pipeline, a high number of products should follow.

GLOSSARY Drug nanocrystal Particle with a diameter between approximately 5–10 nm up to 1000 nm, composed of a pure drug without any matrix material. Drug nanoparticles can be completely crystalline in nature, partially crystalline, or amorphous, and are produced either by bottom-up technologies or top-bottom technologies.

Drug nanosuspension Nanosuspension composed of drug nanocrystals (cf. nanosuspension). Nanosuspension In the strict sense, a suspension of drug nanocrystals in a dispersion medium. Typically, the suspension is stabilized by surfactants or polymeric stabilizers (stearic stabilization). The dispersion medium can be water, mixtures of water with other water-miscible liquids, or nonaqueous media such as oils, liquid polyethyleneglycol (PEG).

REFERENCES 1. E. Merisko-Liversidge, “Particles 2002,” Orlando, FL, Abstract 45, 2002, p. 49. 2. A. T. Florence and D. Attwood, “Physichochemical Principles of Pharmacy.” Chapman and Hall, New York, 1981. 3. D. J. W. Grant and H. G. Brittain, in “Physical Characterization of Pharmaceutical Solids” (H. G. Britain, Ed.). New York: Marcel Dekker, 1995. 4. R. H. Müller, K. Krause, and S. Anger (publication in preparation). 5. http://physchem.ox.ac.uk/MSDS/OL/oleic_acid.html. 6. “The Merck Index,” 12th ed. Merck & Co., Inc., Whitehouse Station, NJ, 1996. 7. “Remington,” 15th ed. Mack Publishing, Easton, PA, 1975. 8. M. Mosharraf and C. Nyström, Int. J. Pharm. 122, 35 (1995). 9. C. Nyström and M. Bisrat, Int. J. Pharm. 47, 223 (1988). 10. R. H. Müller, S. Benita, and B. H. L. Böhm, Eds., “Emulsions and Nanosuspensions for the Formulation of Poorly Soluble Drugs.” Medpharm Scientific Publishers, Stuttgart, 1998. 11. R. Voigt, “Lehrbuch der Pharmazeutischen Technologie.” Verlag Chemie, Berlin, 1984. 12. H. Kala, U. Haak, U. Wenzel, G. Zessin, and P. Pollandt, Pharmazie 42, 524 (1987). 13. “Hagers Handbuch der Pharmazeutischen Praxis.” Springer-Verlag, Berlin, 1925. 14. M. List and H. Sucker, Patent GB 2200048, 1988. 15. P. Gassmann, M. List, A. Schweitzer, and H. Sucker, Eur. J. Pharm. Biopharm. 40, 64 (1994). 16. R. H. Müller and G. E. Hildebrand, Eds., “Pharmazeutische Technologie: Moderne Arzneiformen 2. Aufl. Wissenschaftliche.” Verlagsgesellschaft mbH, Stuttgart, 1998. 17. P. Gaßmann and H. Sucker, GB Patent 2200048; GB Patent 2269536, 1994. 18. P. H. List, “Arzneiformenlehre, p. 205.” Wissenschaftliche, Verlagsgesellschaft GmbH, Stuttgart, 1982. 19. M. R. Violante and H. W. Fischer, U.S. Patent 4, 997, 454, 1991. 20. S. Schenderlein, M. Husmann, M. Lück, H. Lindler, and B. W. Müller, “Proceedings of the 13th Symposium on Microencapsulation.” Angers, France, Sept. 7 2001, p-074. 21. B. W. Müller, EP 0 605 933, 1998. 22. N. Rasenack and B. W. Müller, Pharm. Res. 19, 1894 (2002). 23. R. Buscall and R. H. Otewill, pp. 141–217. Elsevier Applied Science Publishers, London, 1986 24. R. H. Müller, K. Peters, and D. Craig, “International Symposium on Control. Rel. Bioact. Material,” 1996, Vol. 23, pp. 925–926. 25. G. E. Hildebrand, J. Tack, and S. Harnisch, Patent WO 072955A1, 2000. 26. R. H. Müller and B. H. L. Böhm, “Dispersion Techniques for Laboratory and Industrial Scale Processing.” Wissenschaftliche Verlagsgesellschaft mbH, Stuttgart, 2001. 27. K. Schalper, S. Harnisch, R. H. Müller and G. E. Hildebrand, “4th World Meeting on Pharmaceutics, Biopharmaceutics and Pharmaceutical Technology (APV, APGI),” Florence, 2002. 28. K. Schalper, Herstellen hochdisperser Systemen mit Kontunierlichen Mikromischern, Ph.D. Dissertation, Dept. of Pharm.

638

29. 30. 31. 32.

33. 34. 35. 36. 37. 38.

39. 40. 41. 42. 43.

44.

Technology, Biotechnology & Quality Management, Free University of Berlin, 2003. K. J. Illig, R. L. Mueller, K. D. Ostrander, and R. Jon, Pharm. Technol. (Oct. 1996). R. H. Müller, R. Becker, B. Kruss, and K. Peters, U.S. Patent 5, 858, 410, 1999. G. G. Liversidge, K. C. Cundy, J. F. Bishop, and D. A. Czekai, U.S. Patent 5,145,684, 1992. S. Buchmann, W. Fischli, F. P. Thiel, R. Alex, and Mainz, “Proceedings of the Annual Congress of the International Association for Pharmaceutical Technology (APV),” 1994, p. 124. R. H. Müller and K. Peters, “International Symposium on Control. Rel. Bioact. Materials,” 1997, Vol. 24, pp. 863–864. G. Vergnault, “Workshop ‘Nanobiotech Medicine,’ Nanobiotech 2001,” Münster, 2001. M. R. Hilborn, “Particles 2002,” Orlando, FL, 2002, Abstract 48, p. 50. P. Indu and S. Ulagaraj, U.S. Patent 5,922,355. M. Freeman, New Drugs 2, 13 (2001). B. H. L. Böhm, M. J. Grau, G. E. Hildebrand, A. F. Thünemann, and R. H. Müller, “International Symposium on Control. Rel. Bioact. Materials,” 1998, Vol. 25, pp. 956–957. Company brochure, Niro-Soavi Deutschland, 2000. H. Stricker, “(Hrsg.), Physikalische Pharmazie,” Wissenschaftliche Verlagsgesellschaft GmbH, Stuttgart, Auflage, 1987. K. P. Krause, O. Kayser, K. Mäder, R. Gust, and R. H. Müller, Int. J. Pharm. 196, 169 (2000). K. Krause and R. H. Müller, Int. J. Pharm. 214, 21 (2001). R. H. Müller, K. Peters, R. Becker, and B. Kruss, “1st World Meeting on Pharmaceutics, Biopharmaceutics and Pharmaceutical Biotechnology (APGI/APV),” Budapest, 1995, pp. 491–492. K. Peters and R. H. Müller, “Proceedings of the European Symposium on Formulation of Poorly Available Drugs for Oral Administration, APGI/APV,” Paris, 1996, pp. 330–333.

Drug Nanocrystals of Poorly Soluble Drugs 45. B. Böhm and R. H. Müller, Pharm. Sci. Technol. Today 2, 336 (1999). 46. R. H. Müller, B. H. L. Böhm, and M. J. Grau, Pharm. Ind. 61, 74 (1999). 47. R. H. Müller, K. Mäder, and K. Krause, PCT Application PCT/EP00/06535, 2000. 48. R. H. Müller and K. Peters, Int. J. Pharm. 160, 229 (1998). 49. C. Jacobs, O. Kayser, and R. H. Müller, Int. J. Pharm. 196, 161 (2000). 50. B. H. L. Böhm, D. Behnke, and R. H. Müller, “2nd World Meeting on Pharmaceutics, Biopharmaceutics and Pharmaceutical Biotechnology (APGI/APV),” Paris, 1998, pp. 621–622. 51. M. J. Grau, O. Kayser, and R. H. Müller, Int. J. Pharm. 196, 155 (2000). 52. R. H. Müller, C. Jacobs, and O. Kayser, Adv. Drug Del. Rev. (ADDR) 47, 3 (2001). 53. C. Jacobs and R. H. Müller, Pharm. Res. 19, 189 (2002). 54. R. H. Müller and C. Jacobs, Int. J. Pharm. 237, 151 (2002). 55. A. Akkar and R. H. Müller, “Access of Therapeutics to the Brain, 8th UKICRS Symposium,” Belfast, 2003, Abstract Book, Poster 12. 56. K. Krause and R. H. Müller, Int. J. Pharm. 223, 89 (2001). 57. B. W. Müller, U.S. Patent application 08/531, 283, 20.09, 1995. 58. K. P. Krause and R. H. Müller, “28th International Symposium on CRS,” San Diego, 2001, p. 722. 59. J. E. Kipp et al., U.S. Patent application 20020168402 A1, Nov. 14, 2002. 60. F. Sylvan, J. E. Lofroth, and B. Levon, U.S. Patent 5,780,062. 61. B. H. L. Böhm, “Herstellung und Characterisierung von Nanosuspensionen als neue Arzneiform für Arzneistoffe mit geringer Bioverfügbarkeit,” Ph.D. Dissertation, Dept. of Pharm. Technology, Biotechnology & Quality Management, Free University of Berlin, 1999.

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Dye/Inorganic Nanocomposites A. Lerf Walther-Meissner-Institut für Tieftemperaturforschung, Garching, Germany

ˇ P. Capková Charles University, Prague, Czech Republic

CONTENTS 1. Introduction 2. Dye/Inorganic Composites 3. Host Lattices and Guest Species 4. Preparation of Dye–Inorganic Host Composites 5. Structure and Structure Simulation 6. Photophysical Behavior and Potential Applications 7. Conclusions and Future Perspectives Glossary References

1. INTRODUCTION Pigments based on inorganic framework structures and intercalated species have been known for a long time and used by mankind for centuries: the semiprecious stone lapis lazuli and Mayan blue. The first one is a natural occurring mineral, which can be used for the preparation of a valuable and very expensive pigment. Its application in ancient Egypt is questioned [1], but the original stone has been used in many art objects for more than 5000 years (an example is shown in Fig. 1) [2]. The origin of the blue color was solved only recently [2]. The main component is a framework silicate based on the sodalite-type cage. The colored species is very probably the S− 3 radical anion [3]. The second material mentioned has been used as a pigment for murals and other art objects (an example is shown in Fig. 2 [4, 5]). It is of purely artificial origin and its application goes back in pre-Columbian times [4–6]. The blue color seems to be caused by the intercalation of indigo in the channel-type clay ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

mineral palygorskite and/or sepiolite [4–6]. The brilliance of the pigment could be influenced by metal nanoparticles deposited on the pigment during preparation [6]. The most recent interest in dye/inorganic composites has been devoted to the optic properties of the dyes and their changes during adsorption/intercalation on inorganic materials. Apart from the possible applications in photocatalysis, photoelectrochemistry, nonlinear optics, and optoelectronics the photophysical properties of the dyes can be easily monitored by ultraviolet-visible (UV-vis) spectroscopy and allow some insight in the guest–guest interactions under the conditions of constrained microenvironments in the interlayer space of layered hosts or in the mesopores of three-dimensional open frameworks as in zeolites or in other meso- and microporous materials [7–19]. A prerequisite for application of UV-vis spectroscopy is that the solids are transparent for UV or visible light. Nearly all materials discussed in the following are electric insulators with a rather big bandgap, thus fulfilling this condition. An enormous amount of material has been collected in the last 20 years in this field of research and the interest in dye/inorganic composites is still increasing. In order to cope with this huge amount of literature we will cite as much as possible previous review articles [7–19]. Original papers are only cited if they appeared after the excellent review of Ogawa and Kuroda [14] or if we need it for a peculiar argumentation.

2. DYE/INORGANIC COMPOSITES Adsorption and/or intercalation of organic molecules or molecular cations is nearly as old as the field of intercalation; thus it dates back to the 1930s [20, 21]. At this time intercalation of organic compounds was restricted to clay minerals. The adsorption of various organic substances by natural and chemically altered or heat-treated clays can produce color changes in the clay. Well pronounced is the

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (639–693)

Dye/Inorganic Nanocomposites

640

Figure 1. Photograph of an art piece containing lapiz lazuli (bracelet with a scarabaeus, piece from the buriel chamber of Tutenchamun). Photography provided by Dr. D. Wildung [2].

coloration of clays, which have been loaded with aromatic amines. The coloration is caused by oxidation with oxygen and/or by structural iron ions after intercalation via cation exchange [22–26]. In [22] an exhaustive list of color forming amines is given. However, only the color reaction of benzidine has been studied in greater detail [24–26]. The formation of crystal violet was observed as early as 1941, if clays have been treated with dimethylaniline vapor [23]. Colored complexes have also been obtained by the adsorption of certain aromatic molecules on montmorillonites saturated with transition metals in the interlayer space [27, 28]. A further method to get coloration is by an acid–base reaction in which the natural or the acid-treated clay behaves as an acid and the adsorbed molecules, known as an acid–base indicator, accept protons and change their color [24, 29]. Without any chemical reaction clays became colored when loaded with organic dye molecule cations or colored metal chelate complexes via ion exchange. Staining of clays with dyes has been used for the determination of the cation exchange capacity, monitoring the surface acidity, colloidal properties,

Figure 2. Mural painting from Cacaxtla: The blue color is made from the so-called Mayan blue; it is a intercalation complex of indigo in the framework clay mineral palygorskite [4–6]. Reprinted with permission from [4], C. Reyes-Valerio, http://www.azulmaya.com.

and redox active sites. These types of application have been reviewed several times and shall not be discussed in detail here [20, 21, 30–33]. For aniline and some aromatic heterocycles the color reaction due to oxidation is only an intermediate state toward the final formation of conducting polymers. The transformation of aniline to polyaniline can be achieved by the external acting oxidant sodium peroxodisulfate (e.g., in case of a zeolite [33]) or by treating Cu2+ or Fe3+ loaded clays and zeolites with aniline, pyrrol, or thiophene vapors [34–41]. The second wave of interest on the intercalation of colored/or color forming species was devoted to those dye systems, which are photosensitive and/or redox active. It started soon after 1975 when Balzani et al. [42] proposed several cyclic reactions involving transition metal complexes, which would result in the net photodecomposition of water either to H2 , O2 , or both (please note the temporal coincidence with the first oil market crisis!). The key compound in constructing redox cycles for photodecomposition of water became the [Ru(bpy)3 ]2+ complex because of several attractive features [43–45]. The principal design of such a watersplitting redox cycle is as follows [46–49]: D + h −→ D∗

D = dye sensitive to visible light = donor D∗ + A −→ Dox + Ared

1 1 Dox + OH− −→ D + O2 + H2 O 4 2 1 Ared + H+ −→ A + H2 2

(1) (2) (3) (4)

Oxidation and reduction catalysts are necessary for steps (3) and (4) in order to reduce the overpotentials of these reactions. In Lehn’s experiments Pt and RuO2 played this role [46, 47]. It became clear very soon that systems based on this reaction sequence have little chance to be effective unless one designs their architecture at a supramolecular level in such a way to promote the electron transfer from D∗ to A, to separate Dox and Ared to avoid recombination, to separate catalytic sites for H2 and O2 [49, 50]. A crucial step in these redox cycles is the electron flow from the excited sensitizer to the first electron acceptor. First, metal complexes have been used for this purpose [46–50]. Also semiconductors have been proposed as a way of achieving charge separation and have led to molecularly sensitized cyclic water cleavage systems [51]. Nijs et al. introduced clay minerals to solve the “architectural” problem. As electronically insulating but negatively charged solids they can support the catalysts and control the position of the ionic species in the medium [49]. About 1981 the methyl viologen system was introduced in the redox cycles as the first electron acceptor of the excited electrons of the sensitizers, maybe due to its fortunate redox potential and its high stability against further chemical modifications [52]. Later on the search for suitable, more complex architectures of redox cycles for the photodecomposition of water has been extended to further intercalating layered and framework structures (see Sections 3 and 6.3). The previous reaction cycle is the shortest functional cyclic redox system (Fig. 3b, middle and bottom), which can

Dye/Inorganic Nanocomposites

641

(a) –

non-cyclic electron transport – –

cyclic electron transport



±

2 Photons

+

2 Photons

Chl aI (P 700)

pigmentsystem I

+ +

+

Chl aII (P 680)

pigmentsystem II

(b) 2OH

DIIox

DI

Chla II Chla II*

A Ired

Chla I+

QA–

DII*

DII

O2

Diox

A IIred

Chla II+



FeSABx+

+

DI*

Chla I Chla I*

QA

2H

AII

FeSAB(x+1)+

NADPH 2

AI





2OH



Dox

A red

Ru(III)

MV+

D

O2

2H +

D*

Ru(II) Ru(II)*

MV

2+

H2

A hν

Figure 3. Light driven redox reaction cycles in plant photosynthesis and artificial water splitting system. (a) Electron transfer chains in plant photosynthesis [53] (top) and the arrangement of pigments in the reaction center of the photosynthesis unit of Rhodopseudomonas sphaeroides [54] (bottom). (b) Simplified representation of the redox system shown in (a), top, using the terminology of Eqs. (1)–(4) (top), cyclic representation of the equations applied to an artificial photochemical system (middle), and a sketch showing the component assembly in the mixed colloid system which is able to split water under illumination [49].

be used to mimic the light driven reaction sequence of plant photosynthesis. Neglecting that the electron transfer steps in plant photosynthesis (Fig. 3a) are much more complex than those given in Eqs. (1)–(4) one will realize immediately that there are two coupled simple redox cycles [cf. Fig. 3a (top) and Fig. 3b (top)]. The first one is connected with the photosystem (PS) I and the second with PS II. The electron of the excited sensitizer of PS I (ChlaI -P700) is transferred through some intermediate steps to the first stable electron acceptor A—an Fe–S protein—and finally used for the production of reducing equivalents NADPH2 (instead of H2 . This process corresponds with Eq. (4). The oxidized state of the donor (ChlaI -P700+ = Dox  is brought back to the initial state D = ChlaI -P700 by the electron transfer from the first stable electron acceptor of the PS II and not directly from water; this reduced acceptor is a reduced quinone. It is reduced due to the electron transfer from the excited state of the ChlaII P680 = DII+ belonging to PS II. The oxidized state of P680+ gets its electron back from water under oxygen deliberation. Figure 3a (bottom) shows the photosynthetic reaction center of Rhodopseudomonas sphaeroides, which is very similar to the reaction center of the LHCII of plants [53, 54]. It is the place for charge separation and its proper function depends crucially on the right architecture and spacing of the various antenna and electron acceptor molecules. Compared with this the solid-state model for reaction sequence Eqs. (1)–(4) is rather “primitive” (Fig. 3b, middle and bottom). However, it is really astonishing that it works so well, even if it did not come to a technical application. Despite the disappointing result of the efforts to mimic photosynthesis it remains a valuable “by-product” of these trials that the photophysical investigation of various light sensitive and redox active systems brought insight in the organization of complex molecules on the external and internal surfaces of intercalating hosts. In addition some of the dyes investigated found applications as electron mediators in photocatalytic processes and clay-modified electrodes [55, 56]. A new boost to the field of dye-inorganic composites comes from the demand of functional supramolecular arrangements in the strongly growing nanosciences. This boost is also accompanied with a shift of interest, from photocatalysis to lasing, second harmonic generation (nonlinear optics), spectral hole burning, quantum confinement, and quantum superstructures. Before we get into the details of these new developments we would like to remind the reader about early trials to produce supramolecular functional units by a sort of self-assembly. Kuhn used in the 1960s the long-known Langmuir– Blodgett technique [57–59] to produce organized arrangements of molecules, which show properties depending on this arrangement [60–63]. In his seminal article written in 1965 he formulated the principles of supramolecular chemistry [60]. Since his original text is so well written and clear we cannot resist repeating here his original version [60]: Die im folgenden zu beschreibenden Versuche gehen von der Zielsetzung aus, einfachste organisierte Systeme von Molekülen zu bauen, das heißt Anordnungen einzelner Moleküle, welche Eigenschaften haben, die durch die Besonderheit der Anordnung bedingt sind. Während die Moleküle schon in einfachsten biologischen Strukturen in hoch organisierter Weise assoziiert sind,

Dye/Inorganic Nanocomposites

642 fehlt es dem Chemiker an Methoden, um Moleküle in vorausgeplanter Weise aneinanderzufügen. Er kann sehr komplizierte Moleküle synthetisieren, aber er kann nicht wie die Natur Aggregate von verschiedenen Molekülen in geplanter Ordnung bauen. Man kann das ferne Ziel der Herstellung größerer organisierter Assoziate von Molekülen in zwei Aspekten sehen. Einerseits wird man bestrebt sein, einfache biologische Strukturen nachzuahmen; andererseits wird man sich bemühen, ganz unabhängig vom Vorbild der Natur irgendwelche der Phantasie entspringende Anordnungen von Molekülen herzustellen, die nützliche, von der genauen Anordnung der Moleküle im Aggregat abhängige Eigenschaften haben und die somit Werkzeuge mit molekular dimensionierten Bauelementen darstellen.

Kuhn and co-workers studied the photochemical properties of synthetic dye molecules (mainly cyanins) embedded in membrane-like multilayer assemblies. These dye molecules carry long-chain alkyl residues, which fit them in the membrane forming amphiphiles. The authors studied mainly the influence of the distance between the sensitizer molecules and the acceptors on the fluorescence and phosphorescence of the prime photon acceptors. A typical arrangement is shown in Figure 4. It shows also a schematic view of the Langmuir trough, the equipment used for preparation of these assemblies. From the historical point of view it is of interest to give also the sketch of Pockels showing her assembly for measuring dense packing of monolayers. Figure 4d is a microscope view of a ferritin assembly made by the same technique and showing a highly ordered hexagonal arrangement of the iron oxide cores. It corresponds very well with recent arrangements of quantum dots [64]. These studies have been continued by many other scientists in different fields of applications, like nonlinear optics, piezoelectric devices, photoelectron transfer processes, photovoltaic effects, electrochromic, thermochromic, and photochromic devices, field effect devices, and various types of sensors [65, 66]. As far as we know real applications of these systems have not been achieved up to date, maybe due to problems with long-term stability of these molecular assemblies, especially in aqueous media. This may be one of the reasons the assembly of metalorganic or organic dye species within inorganic crystalline host materials by intercalation processes has been investigated. These inorganic hosts (mainly two-dimensional anisotropic or zeolite-type materials) may lead to an increased stability of the organic species, keep them separated from other components in the surrounding medium, and may support preorientation of the guest molecules which could be of advantage for various applications. However, the uptake of bulky organic species in natural or synthetic preformed hosts is rather limited and their orientation is often dominated by the interaction of the -electron systems with the inorganic intracrystalline surfaces, thus preventing guest–guest interactions. On the other hand the question arises whether the arrangement of the dye molecules in the interlayer space is really of importance for electrochemical and photofunctional applications. Cyclovoltaic studies of clay modified electrodes gave some evidence that the capture of the redox mediators in the interlayer galleries prevents the diffusion in the electrolyte medium, but the amount of electroactive species is very low; thus only those species very near the outer surface of the crystals are really involved in the electrochemical processes

(a)

(c)

yellow

c

blue

dark

d

b a

e

(b)

A S

2 10 9

1 8 7 12 6

11

Ultraviolet Light (d) 2

5

4

3

R

H C

R

N+

N

C18H37

C18H37

R=O R=S

(e)

Figure 4. Langmuir–Blodgett technique: (a) Drawing by A. Pockels showing her film balance. Reprinted with permission from [?], A. Pockels. (b) Sketch of a modern Langmuir–Blodgett trough [65]. (c) Cut through a Langmuir–Blodgett assembly made of cyanins shown in (d); the assembly is irradiated by UV light which is absorbed by the sensitizer S. Zone 1, energy transfer from S to the acceptor A at a distance of 50 Å, yellow fluorescence; zone 2, distance S to A of 150 Å, no energy transfer, blue fluorescence of sensitizer S; zone 3, sensitier S omitted, no fluorescence because the light is not absorbed. If irradiated with blue visible light, which is absorbed by A all three compartments show yellow fluorescence of A. Reprinted with permission from [62], H. Kuhn and D. Möbius, Angew. Chem. 83, 672 (1971); Int. Ed. Eng. 10, 620 (1971). © 1971, VCH, Weinheim. (d) Structural formula of the cyanins used to build the Langmuir– Blodgett assembly shown in (c). In the sensitizer S oxygen is the second heteroatom in the five-membered ring (R = O). In the acceptor A the heteroatom R corresponds to sulfur. (e) Ferritin assembly adsorbed on a stearicacidmethylester/trimethyldodeclammoniumsulfate film. Dense packing of the proteins as detected by the ferritin cores in an electron microscope view [62].

[56, 67, 68]. Photophysical investigations have been carried out mainly on dye molecules adsorbed to clay or Zr– phosphate colloids prior to precipitation [14, 15]. Thus the loading with the dyes is below the cation exchange capacity. In these studies metachromic shifts of the absorption bands have been observed which can be interpreted in terms of dye aggregate formation [14, 15, 19, 32]. Since the space limitations in the interlayer galleries prevents mainly aggregate formation several authors conclude that these aggregates are formed on the external surfaces of the colloids or in the mesopores within the tactoid aggregates of partially restacked layers of the host colloidal layers [14, 15]. Apart from these problems of dye distribution in intra-crystalline and external surfaces there is only information about the layer distances which gives some idea about whether we have dye molecules in the interlayer space or not.

Dye/Inorganic Nanocomposites

In case of co-adsorption/co-intercalation of more than one dye species, which is of interest for the formation of photofunctional units, the situation is even worse (see Section 6.2). From the aforementioned problems one can conclude that the distribution of both species in the intracrystalline and external surfaces may be more or less random unless there is a preferred aggregation of AB pairs with respect to AA or BB interactions, or vice versa. Even if there is such a preferred aggregation it is not quite certain that such AB pairs will be unchanged if they are inserted in the interlayer space, though the chance to plan the arrangement of sensitizers and photon or electron acceptors is much lower than in the case of the Langmuir–Blodgett multilayers. In the inorganic/organic composites there is not only the problem of planned distribution of different functional units but also the problem of orientation of the dye loaded solid particles. Due to the uniform size of the membrane forming amphiphiles the dense packing of the molecules is possible under the conditions of the Langmuir–Blodgett technique, even in a macroscopic area. In case of the inorganic particles one has a particle size distribution, which does not have be very regular. In case of anisotropic solids with a fortunate aspect ratio the orientation of the particles parallel to the layers should be without problems, at least in the proper electrolyte medium. However, the orientation within the layers is nearly impossible if there is a particle distribution. It will be hard also to have a dense packing of the particles without any macroscopic voids. In order to overcome all these problems completely new preparation procedures have been developed: layer-bylayer deposition [19, 69–72] and the in-situ synthesis of the organic/inorganic composites [19, 73–75]. The prerequisite for both techniques is a new class of dye molecule derivative which shows great potential to form stable well-ordered assemblies on an inorganic solid template and in aqueous solutions, respectively. In almost all cases they consist of the aromatic dye unit with an elongated shape (azo-benzenes, styryl compounds), with strong charge transfer interactions (porphyrins or phtalocyanines without and with complexing metals) or they are modified in such a manner that they are substituted with long tails of alkyl chains and positively or negatively charged head groups. The chromophor is in the center of the molecule. Some systems used recently are shown in the scheme. Up to now the first technique has been applied to clay minerals, Zr-phosphate, and some layered oxide systems forming new forms of clay modified electrodes [71, 72] or photofunctional units with charge separation after illumination [69]. The second is rather new. Structure directing assemblies containing chromophores have been used as templates in hydrothermal processes, in which the inorganic material is deposited on the surfactant system. This follows the lines of synthesis of the large pore zeolite analogs of MCM type. In this procedure new dye/inorganic composites have been obtained with the following inorganic matrices: M41S [73, 74] and anionic clays [75]. In these samples completely new arrangements of the chromophores have been realized and the packing density seems to be higher than in the normal intercalation compounds. Both methods may help to design the dye organization but do not solve the problem of particle orientation and

643 the formation of micropores. One trial to tackle the problem of particle orientation is the successful directed growth of zeolites with one-dimensional channels [13, 16, 76]. The as-grown crystals are arranged in such a manner that the channels are aligned perpendicular to the solid support. The previous discussion was devoted mainly to the layered dye–inorganic host composites. Since about 1985 there has been growing interest on dye species inserted in framework structures like silica gels, zeolites, and zeolite-like frameworks [8, 13, 16–19]. Due to the diameter of channels or supercages the dye–dye interactions are much more restricted than in the case of layer composites and also the dimension of the inserted dye molecules is restricted by the size of the pores. In many cases only linear arrangements of the long axis of the dye molecules in parallel channels are possible. The rigid walls of the channels prevent large-scale rotations and the diffusion of the molecules. These peculiar and strongly restricted preorientations may be of advantage for applications in quantum superstructures and other optic and electronic devices. This preorientation is visualized in the response to light of preferred polarization as shown in

(a) 50 µm

(b)

Figure 5. Azobenzene loaded AlPO4 -5 crystals between polarizers crossed at an angle of 70 , which gives optimal visual contrast (the directions of polarization are indicated by the crossed arrows). One crystal marked with a pointer was irradiated: before (top) and after (bottom) irradiation. Reprinted with permission from [77], K. Hoffmann et al., Adv. Mat. 9, 567 (1997). © 1997, VCH, Weinheim.

Dye/Inorganic Nanocomposites

644 Figure 5. The special microenvironment in these materials may prevent also the diffusion out of the channel and will lead to a higher chemical stability of the dye molecules. The inclusion of rather big dyes during synthesis of the frameworks will improve the stability and could be of advantage for the application of the dye–zeolite composites as new pigments. The new mesoporous systems with rather big channel diameters are now accessible for a larger number of dyes even those with more bulky shape [18].

3. HOST LATTICES AND GUEST SPECIES 3.1. Layered Host Lattices Layered structures represent a very suitable matrix for insertion of various guest species, because they are able to accommodate very large guest molecules in the interlayer space by the free adjustment of the interlayer space [78–83]. The number of host systems investigated with respect to the uptake of dye molecular species is much more restricted than the number of host systems known to undergo intercalation processes. Those host lattices for which dye intercalation has been described are collected in Table 1 (whereas a comprehensive list of layered host lattices for intercalation is given in [83]). With the exception of 2H-TaS2 all compounds listed in Table 1 are electronic insulators or wide gap semiconductors. The most important layered host materials are the socalled 2:1 clay minerals, natural and synthetic ones. They consist of two flat corner-shared Si–O tetrahedral networks whose apices points toward each other and take part in the formation of the central metal-oxide/hydroxide layer, which

is of the brucite-type structure (see Fig. 6). In the ideal case all the octahedral holes are occupied by Mg (talc) or Al ions (pyrophyllite) only (trioctahedral or dioctahedral, respectively). Normally some of these ions are replaced by other metal ions of the same or different positive charges. Also the Si ions in the tetrahedral layers can be substituted by Al3+ or Fe3+ ions. The substitution by other valent ions leads to an overall negative charge of the clay layers, which has to be compensated by hydrated cations in the interlayer space (not shown in Fig. 6). These interlayer cations (normally alkali metal ions) can be exchanged easily by other inorganic and a huge number of organic cations, including cationic dyes. In addition these interlayer cations undergo very peculiar interactions with neutral solvating species. The well-developed intercalation chemistry of clay minerals is the subject of many review articles, for example [20, 21, 83, 85–87]. The majority of the clays which have been used for dye intercalation belong to the class of smectites characterized by a low negative charge. The most frequently used minerals are the montmorillonites (dioctahedral); others are the trioctahedral minerals hectorite, fluorohectorite (mostly synthetic), saponites, and the synthetic laponite (similar to hectorite) [14, 32]. Applications of higher charged 2:1 clay minerals in dye intercalation are rare and restricted to some vermiculites [32] and the synthetic fluoro-taeniolite [88, 89]. Most of the investigations concerning dye intercalation are carried out in colloidal solutions of the clays because the uptake of the dyes can be controlled much more easily than the ion exchange of complex molecules in crystalline materials [14, 15, 20, 21, 32] and easily monitored by UV-vis spectroscopy. This is the reason montmorillonite dominates in

Table 1. Layered solids used as hosts for dye intercalation reactions. Solid type (i) Uncharged layers Metal dichalcogenides Metal phosphorous chalcogenides Oxides Metal oxyhalides Metal phosphates/phosphonates (ii) Charged layers (a) Positively charged hydroxides [Hydrotalcite type compounds (“anionic clays”)] (b) Negatively charged clays

Layered silicic acids Oxides

Formula of examples (mineral name)

Ref.

2H-TaS2 MPS3 = Mn, Cd

[130–132] [223–225]

V 2 O5 FeOCl -Zr(HPO4 2 -Zr(RPO3 2 ; R= organic residue

[122–127] [128, 129]

[MII1−x MIII x (OH)2 ][Ax/n ]mH2 O MII = Mg, Fe, Co, Ni, Mn, Zn; MIII = Al, Fe, Cr, Mn, V; An = inorganic or organic anions Nax (Al2−x Mgx (Si4 O10 (OH)2 (montmorillonite) Cax/2 (Mg3 (Alx Si4−x O10 (OH)2 (saponite) M(I)0 67 (Mg2 65 Li0 34 (Si4 O10 (OH,F)2 (hectorite) Li0 8 (Mg2 2 Li0 8 (Si4 O10 (F)2 [fluorohectorite (synth.)] 3+ (Mg)x (Mg3−u Aly Fe2+ v Few  (Si4−z Alz O10 (OH)2 ; u = y + v + w; x = z − y − w (vermiculite) Li(Mg2 Li)(Si4 O10 F2 (taeniolite)

[14, 105–108]

Na2 Si4 O29 · nH2 O (magadiite) Na2 Ti3 O7 , K2 Ti4 O9 K4 Nb6 O17 · 3H2 O K[Ca2 Nan−3 Nbn O3n+1 ], 3 < n < 7

[14] [14] [14] [32] [88, 89] [93] [111–120]

Note: Their structures and intercalation chemistry are described in [78–83]. Literature directly related to dye intercalation is given in the table.

Dye/Inorganic Nanocomposites

645

(a)

(b) Zr P O

a c

Figure 6. Structure of an isolated monolayer in 2:1 clays shown: perpendicular (a) and parallel (b) to the layers. The drawing is made with a Cerius2 modeling unit. The crystallographic parameters adapted from [84].

dye–clay composite research: It can be transferred easily in the colloidal state and its colloidal solutions are well understood and very stable [90–92]. Dyes to intercalate in clay minerals have to be cationic. Very recently the first intercalation of a dye in the layered polymerized silicic acid mineral magadiite has been described [93]. (Please note this silic acid is structurally related to the tetrahedral layers of the clay mineral host lattice [94].) The next important layered host compounds are the layered Zr–phosphates/phosphonates. The -modification of Zr(HPO4 2 · xH2 O, whose structure is shown in Figure 7, is the most important member of this family of compounds with respect to intercalation chemistry. The intercalation chemistry of this host lattice is described several times [96–99]. In Table 1 it is counted as a neutral compound. Its intercalation chemistry is dominated by the presence of weak Bronstedt acid groups. Compounds with functional groups able to accept protons are the preferred guests and for this reason the intercalation of complex species such as the cyclodextrins or porphyrins has been achieved using their amino derivatives. Ion exchange is possible in a medium controlling the pH and is supported by the hydration water or by other species propping open the interlayer space. For example, the half-exchanged phase with the composition ZrNaH(PO4 2 · 5H2 O is formed at a pH > 2, whereas the fully exchanged phase appeared at a pH > 6. In contrast less hydrated large monovalent ions like Cs or highly hydrated divalent ions are not able to replace protons at slightly

Figure 7. Perspective view of structure of -zirconium phosphate, -Zr(HPO4 2 . H2 O (parallel to the layers) prepared with DIAMOND— Visuelles Informationssystem für Kristallstrukturen (Professor Dr. G. Bergerhoff, Bonn, Germany). The red dots in the interlayer space are water molecules. Crystallographic data adapted from [95].

acidic or basic media [99]. They can form also colloidal solutions [88], which can be used for the oriented deposition of organic inorganic composites. Unique for all intercalating systems is the following property of the Zr–phosphate system: If phosphoric acid is replaced by organic phosphonates (the chromophore is part of the organic moiety) and these are brought to reaction with Zr ions [71, 72] a solid is formed which has the same arrangement of the Zr–PO4 units as in the purely inorganic -Zr(HPO4 2 · xH2 O. Thus, the Zr4+ /phosphate/phosphonate system is an ideal candidate for layer-by-layer deposition of inorganic/organic multilayer assemblies. Their structure is not only determined by the bonding restrictions of the Zr–phosphate arrangement but also by the structure directing influence of the organic counterpart. However, it should be noted that the product of this assembly is no longer an intercalation compound but a three-dimensional covalent network with alternating inorganic/organic layers [69, 71, 72]. A very interesting and versatile class of host lattices is the so-called “anionic clays” [100, 101]. Their chemical composition is given by the general formula [(M2+ 1−x (M3+ x (OH)2 ]x+ (Am− x/m · nH2 O where M2+ = Mg, Ni, Fe and M3+ = Al, Cr, Fe. The preferred anion in these compounds is CO2− 3 , which can be replaced easily by other ions, even very complex ones via ion exchange reactions [100, 101]. A strongly related group of intercalation

Dye/Inorganic Nanocomposites

646 compounds can be achieved, if gibbsite is treated with an excess of Li salts [102–104]. The resulting composition of this reaction is [LiAl2 (OH)6 ]X · nH2 O with X = Cl, Br, NO− 3 . The structure of both series of compounds consists of a brucite-type hydroxide layer structure and the anions are inserted between these hydroxide layers. In case of the gibbsite the Li ions are inserted in the empty octahedral lattice sites of the Al(OH)3 /gibbsite layers. Figure 8 shows this structure for the example [LiAl2 (OH)6 ]NO3 . Apart from porphyrin derivatives dye anions have been intercalated in the layered hydroxides only recently [105–108]. Because these compounds can be prepared at relatively moderate temperature it is possible to obtain dye–MOH composites by direct synthesis in the presence of the organic moiety as structure directing medium [75]. In order to be inserted in these compounds the chromophore systems should be modified with functional groups giving them an excess negative charge. There is a large number of layered ternary or quaternary oxides of titanium and mixed systems containing titanium and niobium [14, 109, 110]. The layers consist of Ti/Nboxygen octahedra sharing edges and corners. The most compact slabs are observed for oxides Ax Ti2−x Mx O4 ; they consist of a double plane of edge-sharing octahedra similar to the array observed along the (110) plane of the sodium chloride structure. The layers of the other compounds are deduced from that structure by shearing the layers in the direction perpendicular to its plane every second, third, or

fourth octahedron [109]. Thus the layers are no longer flat but corrugated as shown in Figure 9 for Na2 Ti3 O7 . Very peculiar is the structure of K4 Nb6 O17 · 3H2 O: the corrugated Nb-oxygen layers are stacked in such a manner that two types of alternating interlayers with different reactivities are formed (Fig. 9) [112–114]. The interlayer galleries of these oxides are accessible for guest species only, if the potassium ions have been replaced with hydrated protons [109]. For the intercalation of more complex species it is

a

0-2 Ti+4 Na+1

c

b

a

Al O N Li b

c

Nb K 0

Figure 8. Perspective view of the structure of [LiAl2 (OH)6 ]NO3 , prepared with DIAMOND. The metalhydroxide layers are built by a hexagonal dense packing of OH and metal atoms occupying all octahedral holes. It is isomorphous with the structure of CdJ2 or the TiS2 layers. The nitron atoms (blue balls) are surrounded by six oxygen atoms (red balls) in the interlayer galleries. This is due to the rotation symmetry of the triangular NO−3 . Crystallographic data adapted from [104].

c

Figure 9. Perspective view of the structure of: Na2 Ti3 O7 (top) and K4 Nb6 O17 · 3H2 O (bottom) prepared with DIAMOND. Please note the two kinds of interlayer space which behave differently in intercalation processes. Crystallographic data adapted from [111, 112].

Dye/Inorganic Nanocomposites

of advantage to use organo-ammonium-exchanged materials as intermediates (propping-open method [83]). The number of dye species intercalated, of course cationic species only, is rather small [14]. The reason they found attention in context of photochemistry was their ability to split water [14, 115–118]. The deposition of Pt and/or NiO may support this process [115, 118]. Some of these oxides can form colloidal solutions, which allow the layer-by-layer assembly [72, 119, 120]. The just mentioned binary oxide (V2 O5  is applied in intercalation reactions as a xerogel which exhibits a layered structural organization by stacking of the V2 O5 sheets separated by water layers. The insertion of ions and organic molecules is governed by a variety of interaction mechanisms: hydrogen bonding, protonation, and charge transfer. Its intercalation chemistry has been reviewed by Livage [121]. Apart from the formation of conducting polymers [122, 123] and the intercalation of benzidine [124] and the electron donor tetrathiafulvalene [125] the intercalation of dye species is rather new [126, 127]. The other uncharged host lattices mentioned in Table 1 play a minor role in the intercalation of dye species. The metal phosphorous chalcogenides [78] are known to intercalate just the hemicyanine/stilbazolium and the [Ru(bpy)3 ]2+ complex [14]. The intercalation of cationic species in this host lattice is accompanied by a loss of structural metal atoms. To the best of our knowledge FeOCl is known to intercalate TTF derivatives and no other dye species [128, 129]. The uptake of this donor leads to a reduction of the host iron atoms. It cannot be excluded that further dye species can be intercalated in these host lattices. 2H-TaS2 is the only metal-like conductor in the compounds listed in Table 1. Up to now only the intercalation of crystal violet [130], methyl viologen [131], and methylene blue [132] has been demonstrated. Because the sample is an electronic conductor the intercalation can be carried out electrochemically. It is interesting to note that in case of methylene blue the packing density and the arrangement of the dye molecules can be influenced depending on the reaction conditions [132] (see also Section 4). The resulting products are superconductors with rather high Tc values in comparison to other organic intercalates [131, 133].

3.2. Framework Structures The framework structures used for the formation of dye/inorganic nanocomposites are collected in Table 2. Most of these frameworks are based on the main the group elements Si, Al, P. These elements are surrounded by four oxygen ions in tetrahedral coordination and these tetrahedra share corners with neighboring tetrahedra. Due to the kind of linking of the tetrahedra a huge number of amorphous and crystalline phases can be imagined and many of them are realized [134]. The variety based on Si only and Si/Al is enormous. Whereas in case of the silica gels the occluded wholes vary in size and the linkage of the tetrahedra is more or less random, the crystalline phases show very regular distribution of cages, supercages, and channels depending on the preparation conditions and the size of the template used during synthesis. In case of the pure silica systems the species formed are the so-called chrathrasils

647 [135]. In order to insert dye species they have to be applied as templates during synthesis. These systems are inclusion compounds where the structure of the host is directed from the included guest and the guest cannot leave after synthesis. Only some metal complexes with interesting photofunctions have been inserted in such sort of frameworks [16, 136]. Replacing part of the silicon atoms by aluminum lead to the huge family of natural and synthetic zeolites [134, 137]. Due to this replacement the framework carries negative charges, which has to be compensated by cations in the cages. The Si–Al–O networks in the zeolites are built in such a way that two- or three-dimensionally intersecting channels allow exchange with chemical species in the surrounding of the solids. At the intersection of channel systems large supercages appear which allow the in-situ synthesis (“ship in a bottle” strategy) of rather complex molecules from smaller building units. The size of the species, which can be inserted from the outside, is, however, restricted by the smallest dimension of the channels. The channel width of most natural zeolites is normally too small for the uptake of the rather big dye species. Therefore, the zeolites used for dye insertion are the synthetic variants with pore size in the order of 7 Å. Table 2 gives a collection of them with the relevant channel and cage dimensions. A frequently used framework is ZSM 5 whose structure is shown in Figure 10. Isostructural to this framework is silicalite, the end-member of the ZSM family without any Al in the structure [139]. In 1982 a new class of microporous framework was discovered replacing Si by Al and P [140]. The compounds are isoelectronic to SiO2 and the overall structure is dominated from tetrahedral networks [134, 141]. However, new theoretically possible structures are realized and some of them have rather big channel diameters of about 11 Å (Table 1). In all the systems of this family (AlPO4 -5, AlPO4 -11, VPI-5) used in the preparation of dye–inorganic nanocomposites the channels are arranged parallel to each other as shown in Figure 11. It is possible to replace a small fraction of aluminum by Si, Fe, or other transition metals. In case of Si the framework again gets a negative charge, which is compensated by cations in the channels [146–148]. The structure seems to be unchanged. Up to now mainly those systems have been used for dye insertion, which has the same structure as AlPO4 -5 [149, 150]. A new family of molecular sieves (M41S) with an open pore diameter of 15 to 100 Å has been described in 1992 [151, 152]. They are based on silica and are grown in the presence of amphiphilic ammonium ions as templates, which are applied in such a concentration that they form hexagonal cylinders. The resulting silica solids are arranged in a honeycomb packing of channels with an open pore diameter of 15 to 100 nm. The inorganic matrix is stable after removal of the template at least to temperatures up to 500  C. The empty channel can be filled easily with other organic species. In these large pore systems there is nearly no restriction of the shape of the guests. In addition rather complex dye species can be dissolved in the surfactants and, thus, inserted during synthesis [19, 153–156, 178, 179]. Using the method applied to the synthesis of phthalocyanine encapsulated in MCM 41 Zhou et al. could get similar mesoporous dye-loaded systems with VOx , FeOx , MoO3 ,

Dye/Inorganic Nanocomposites

648 Table 2. Framework structures [10, 134].

Compound

Effective window size (Å)

Pore shape and size (Å)

Dimensionality

Sodalite Zeolite A Zeolites X, Y ZSM-5

2.3 4.2 7.4 5 3 × 5 6 5 1 × 5 5

cage (6.6) cage (11.4) cage (11.8) channel

three three three two, intersecting

Mordenite

7 0 × 6 7 2 6 × 5 7

channel

two, intersecting

Zeolite L Silicalite-1

7.1 5.5

AlPO4 -5 AlPO4 -11 VPI-5 [215] SAPO-5 (AlPO4 -5) SAPO-34 (chabazite) FAPO-5 (AlPO4 -5)

7.3 3 9 × 6 3 12–13 8 4.3 7.3

Nonasil [135, 136] Octadecasil [136] Dodecasil 1H [136]

— — —

lobe (7.5) channel

single three

channel channel channel channel channel channel

single single single single three single

cage (4.1) cage (4.05) cage (4.7)

zero zero zero

MCM 41 MCM 50 Sepilolite Palygorskite

3 7 × 10 6 3 7 × 6 4

and WO3 frameworks [154–156]. However, the metal coordination and the formation mechanisms of these frameworks are not clearly established [157]. Other framework structures with parallel running channels are palygorskite and sepiolite (see Fig. 12). They are considered phyllosilicates because they contain a continuous two-dimensional tetrahedral network of composition Si2 O5 , but they differ from the layered 2:1 aluminosilicates in lacking continuous octahedral sheets [158]. The structures of these minerals can be considered as consisting of 2:1 phyllosilicate ribbons, one ribbon being linked to the next by inversion of the SiO4 tetrahedra along a set of Si–O–Si bonds. The channels thus created are running parallel to the

fiber axis. The intercalation chemistry of these minerals is still in its infancy, but it got some attention by the discovery that the ancient pigment Maya blue can be considered an intercalation compound of indigo into palygorskite [4–6]; the proposed structure is shown in Figure 12 (bottom). Only recently has the intercalation compound of indigo in sepiolite been studied in greater detail [159].

Al O

b a

Faujasite

b a

T O

c

a b

Al P O H

P Al O

a

Figure 10. Structure of zeolites. Right: supercage assembly of faujasite structure (top) and possible position of sodium ions [216]; the synthetic zeolites X and Y have the same structure as the natural counterpart faujasite. Left: view of the ZSM-5 structure along the crystallographic b-axis showing the more regular channel windows. This view is prepared with DIAMOND. Crystallographic data adapted from [138].

Figure 11. Structures of various AlPO4 compounds used in the preparation of dye–inorganic nanocomposites: The view shows the arrangement and shape of the parallel channels in these compounds. The viewgraphs are prepared with DIAMOND. Crystallographic data adapted from [142–145].

Dye/Inorganic Nanocomposites

649

c sin β = 12.7 Å

b = 17.9 Å

PALYGORSKITE

c = 13.37 Å

b = 26.95 Å

SEPIOLITE +

=O

= OH

= H 2O

= Si = Al & Mg

Figure 12. Structure of sepiolite and/or palygorskite. Top: Ball and stick drawing of the crystallographic structure of the minerals sepiolote and palygorskite. Bottom: Modeling structures of palygorskite (left) and the proposed Mayan Blue structure (right). Reprinted with permission from [4], C. Reyes-Valerio, http://www.azulmaya.com.

3.3. Guest Species Species of nearly any class of organic dyes have been inserted in inorganic solids: acridine, xanthen, phenazine, phenoxazine, and phenothiazine type compounds, porphyrins and phthalocyanins, and only recently azocompounds and cyanine derivatives. In addition there are some metal compounds which belong in the context of this chapter. Three major phenomena induced by light absorption are discussed with respect to possible applications: (a) electron transfer, (b) photochromism [160], and (c) second harmonic generation [161, 162]. These effects are related to the molecular structure of the dyes and can be assigned roughly to different classes of dyes: Compounds inducing electron transfer should be redox active irrespective of the influence of light, the redox reactions involved should be fully reversible, and light absorption may change the redox potential. Compounds fulfilling these conditions are, for example, the metal chelate complexes, porphyrins and phthalocyanines, and the phenothiazines. Photochromism means a reversible transition between two states with different light absorption spectra, where the transition should be induced by light at least in one direction. Known organic photochromic systems are based on cis–trans isomerizations (e.g., in stilbenes, azobenzenes, indigo, or retinal—the visual pigment of our eyes), pericyclic reactions (spiropyranes), on redox reactions (e.g., thionine, viologens). Second harmonic generation (SHG) affords molecules, which show strong dipole moments. Such systems

are benzene or stilbene molecules carrying simultaneously electron-pushing and electron-withdrawing substituents. In order to get a SHG effect the molecules must be incorporated in non-centro-symmetric environments. Since the inorganic matrices used for immobilization or orientation of the organic dyes can be neutral (zeolite related materials like AlPO4 modifications, silicalite) or charged, these dyes are inserted preferentially, which correspond with the charge status of the solid. The overwhelming number of hosts is negatively charged; thus dye cations will be inserted preferentially. There are many dye species which are cationic in nature and thus fulfill this condition. If this is not the case it is always possible to incorporate in the periphery of the chromophore substituents with the right charging of the resulting molecules without influencing too much the absorption behavior. Examples are as follows. (a) Whereas pyridyl or anilinium subsituents make porphyrines more soluble, the uptake in negatively charged zeolites or clay minerals SO− 3 groups allow easier insertion in the positively charged double hydroxide hosts. (b) Methyl orange and methyl yellow are azo-compounds; the first one is neutral or protonated and can be inserted in framework structures and negatively charged hosts and the second one carries an additional SO− 3 group suitable for the uptake in the double hydroxides.

3.3.1. Metal Complexes Although organic dyes are the focus of this chapter, it is worth mentioning a few metal–organic systems (see Table 3). The cobaltocinium cation is the first metal–organic complex which acts as a template in the synthesis of microporous solids. Since the framework used is of the chlathrasil type the fluoride counterion of the metal complex has to be included in the solid in order to keep the overall charge neutrality [136, 163]. This inclusion complex shows an electric field induced SHG effect [164]. Transition metal bipyridine and phenanthroline complexes found widespread interest in clay chemistry because of their optical activity and the potential application in adsorption and asymmetric synthesis [165]. Soon after the proposed value of metal complexes in photoelectrolysis of water [42] tris(2,2′ -bipyridine)ruthenium(II) {[Ru(bpy)2+ 3 ]} became the most important photosensitizer in all artificial Table 3. Colored metal chelate complexes in dye/inorganic nanocomposites [14, 18, 19, 165]. 2,2′ -Bipyridine complexes

1,10-Phenanthroline complexes

[Cr(bpy)3 ]3+ a [Fe(bpy)3 ]2+ [Ru(bpy)3 ]2+ [Cu(bpy)3 ]2+ [Cu(bpy)2 ]+

[Fe(phen)3 ]2+ [Ni(phen)3 ]2+ [Ru(phen)3 ]2+ [Cu(phen)2 ]2+ [Ru(BPS)3 ]4− [167]

Various ligands Co(saloph) [166] Co(salen)

CoIII (Cp)2 F [136, 163, 164]

Note: Underlined references refer to intercalation in framework structures. References behind a species are additional references to those given in the headlines. a bpy = 2,2′ -pipridyl, phen = 1,10-phenanthroline, BPS = 4,7-diphenyl-1,10phenanthrolinedisulfonate, saloph = N ,N -bis(salicylidene)-1,2-phenylendiamine, salen = N ,N ′ -bis(salicylidene)-ethylenediamine, cp = cyclopentadienyl.

Dye/Inorganic Nanocomposites

650 water splitting assemblies [14, 18, 43–52, 69] because of its unique combination of chemical stability, long excited state lifetime, and redox properties. Its photofunctional properties are well characterized and it may be the metal complex, which has been intercalated in as many inorganic hosts as possible. The intercalation of this Ru complex in smectites has been described by Traynor et al. as early as 1978 [168].

3.3.2. Porphyrins and Phthalocyanines The naturally occurring porphyrin chlorophyll is the most important light harvesting antenna in the biosphere [53]. Whereas porphyrin complexes play an essential role in artificial organic supramolecular systems mimicking photosynthesis [51, 194], they could not compete with the Ru complex in the inorganic models [51]. Nevertheless intercalation compounds in clays have been known for many years due to their potential importance in biogeochemistry [169, 170]. Quite interesting with respect to prebiotic chemistry is the formation of phorphins from aldehydes and pyrrols in the interlayer space [171]. In most cases synthetically modi-

fied porphyrins has been used carrying charged substituents CH groups (Table 4 and Scheme 1). The posion the tive charge can be increased further by proton uptake of some of the pyrrol units [170–174]. Thus, each of the porphyrin molecules can compensate the charge of at least four cations and thus it should be enough space for these rather big molecules. Depending on the cation in the interlayer galleries the porphyrins: a) are inserted in the interlayer space; b) remove metal atoms out of the interlayer space and form complexes with them in solution (in case of copper ions); c) form complexes with metals in the interlayer space; or d) are co-intercalated to remaining metals [170–174]. Bathochromic shifts of UV-vis-absorption bands are mainly interpreted as the appearance of different protonated states in the interlayer space as function of water acidity in the presence of certain metal atoms. In most of these cases the porphyrin molecules are laying parallel to the host layers. Only in the case of the exchange of preformed porphyrin–metal complexes (Fe3+ , Co2+ complexes of TAP and TMPyP) into montmorillonite

Table 4. porphyrins and phthalocyanins in dye/inorganic nanocomposites [14, 19, 170–174]. Metal-free systems Hematoporphyrin IX [187] TPP = meso-tetraphenylporphyrin [191] TpyP = meso-tetrapyridylporphyrin [191] TMPyP = meso tetrakis-(1-methyl-4-pyridyl) porphyrin [183, 190, 191] THPyP = meso tetrakis-(1-hexyl-4-pyridyl) porphyrin [19, 180] TAP = TMAP = meso tetrakis-(N ,N ,N -trimethyl-4-anilinium)porphyrin [190]a TAHPP = meso-tetrakis(4-(6-trimethylammino)hexyloxy)phenylporphyrin [73] TTMAPP = meso-tetrakis(5-trimethylamminopentyl)porphyrin [179] POR = 5,15-di(p-thiolphenyl)-10,20-di(p-tolyl)porphyrin [192] TSPP = meso tetrakis(4-sulfophenyl)porphyrin [184, 189] pTCPP = meso tetrakis(4-carboxyphenyl)porphyrin [189] oTCPP = meso tetrakis(2-carboxyphenyl)porphyrin [189] Pc = phthalocyanine [154, 156] TSPc = 2,9,16,23-tetrasulfophthalocyanine [19, 180] TMPyoPc = 2,9,16,23-tetrakis[3-(N -methylpyridyloxy)] phthalocyanine [19, 180, 186] TMPyPaz = tetra[N -methylpyrido(2,3-b:2′ ,3′ -g:2′′ ,3′′ -1:2′′′ ,3′′′ q)]porphyrazine [19, 180] TEPyPaz = tetra[N -ethylpyrido(2,3-b:2′ ,3′ -g:2′′ ,3′′ -1:2′′′ ,3′′′ q)]porphyrazine [19, 180] 15-crown-5-PC = 4,5,4′ ,5′ ,4′′ ,5′′ ,4′′′ ,5′′′ -[tetrakis(1,4,7,10,13-pentaoxatridecamethylene)] Metal-containing systems 4+

3+

Sn –TTP, Fe –TPP, VO–TTP, Mn3+ –TTP, Mg–TTP, Zn–TTP, Cu–TTP Sn4+ –TpyP; Fe3+ –TpyP Fe3+ –TMPyP [190], Co2+ –TMPyP [180], Zn2+ –TMPyP [19, 179, 180]; Cu–TMPyP [182, 190] Zn2+ –THPyP [19, 180] Fe3+ –TMAP [174, 185]; Zn–TMAP [19, 176–178]; Cu–TMAP [179] Cu–TAHPP [73] Chl = chlorophyll [154, 188] ZOR = 5,15-di(p-thiolphenyl)-10,20-di(p-tolyl)porphyrin–Zn [192] Cu–Pc [154–156] Zn–TMPyoPc [19, 180, 186] Cu-tetra-4-dimetylaminophthalocyanine [175] Zn–TEPyPaz [19, 180] Zn–TTMAPP [179] Co–TSPc [181] Cu–TSPc [193] Note: Underlined references refer to intercalation in framework structures; italic references refer to encapsulation in gels. References behind a species are additional references to those given in the headlines. a Different acronyms have been used in the literature.

Dye/Inorganic Nanocomposites

651 3.3.3. Viologens

R

Porphyrins N

N R

H

H N

R N

R N+ CH3

R=

O S O O

R=

CH3 +

R=

N

CH3 CH3

O

R=

(CH2)6

+

N

CH3 CH3 CH3

R

Phthalocyanine

N R N + C2 H5 C 2H5 N+

C2H5

NH N

N N Zn N N

+

N R

HN N

N R = -SO3H

N

N

N

N

R= N

N+ C2H 5

R

O N CH3

N

Zn-Tetra[N-ethylpyrido(2,3-b:2'3'g:2",3"-1:2"',3"'q)]porphyrazine

Scheme 1.

and fluorohectorite are the complexes oriented in upright positions allowing porphyrin–porphyrin interactions [174]. In very recent publications the bathochromic shift of absorption bands is explained by a flattening of the TMPyP molecules due to rotation of the pyridyl side-groups to a coplanar arrangement with respect to the porphyrin ring system and an enhanced bonding to the clay layers [176, 177]. This change in configuration shall lead to the formation of a densely packed monolayer on the external and internal surfaces of the clay particles (“size-matching effect”). The fluorescence lifetime of the porphyrins in this arrangement seems to be sufficiently long to be used as sensitizers [177]. Photochemical energy transfer between porphyrins adsorbed on the same clay sheet and those adsorbed on adjacent sheets has been observed recently [178]. The question arises as to whether these new observations are related to the use of synthetic clays. Whereas porphyrins and phthalocyanines are too big to be inserted in classic zeolites they have been incorporated into micro- and mesostructured materials after synthesis or during synthesis of the mesostructures [19, 153– 156, 179, 180]. With a new porphyrin derivative, which acts as a structure directing amphiphile, a new layered M41S mesostructure has been prepared consisting of an alternative stacking of the porphyrin and the silica layers [73]. Apart from applications as sensitizers, porphyrins and phtalocyanines incorporated in solids could find applications in SHG and information storage [180]. Up to now the hole burning effect has been shown in a Zn–phthalocyanine/AlPO4 -5 system [187].

Viologens form an outstanding group of chromophores (some of them are shown in Scheme 2; those systems interesting in the context of this chapter are collected in Table 5) because their electrochemical behavior is in the center of interest, not their photophysical properties. Easily accessible are only the dicationic RV2+ salts. RV2+ undergoes reversible reduction to the deeply colored monoradical RV+• and in a second one-electron step to the neutral biradical RV•• [195]. Therefore, viologens received much attention as redox indicators in biochemistry, as electron transfer mediators in electrochemistry [55, 56, 195], and as a first electron acceptor in photoinduced redox processes like artificial photosynthesis [18, 51, 69]. This latter application is based on the observation that colloidal Pd led to the disappearance of the MV+• radical (obtained by reduction of MV2+ with CrCl2  with concomitant formation of H2 [197]. This metal catalyzed water splitting with MV as the electron transfer system was applied to light induced processes around 1980 [51]. As a consequence of this success MV2+ got the most studied electron acceptor in photofunctional redox arrangements with Ru(bipyr)2+ 3 as the photosensitizer in the presence of various electron donors [51], even in the presence of clays [15, 198], as well as in corresponding zeolite systems [16–18]. The intercalation of methyl viologen was first studied when it was observed that soil clays inhibit the herbicide effect of methyl viologen (=paraquat) action in agricultural applications [199]. Subsequent studies (reviewed in [87]) led to a detailed model of the MV2+ arrangement in the interlayer space of vermiculites [200]. To the best of our knowledge this could be the first successful structure determination for a dye/inorganic nanocomposite. Because the reduction is accompanied by color formation viologens could be of interest as redox mediated photochromic materials. Only a few publications are devoted to

+

N+

N

+

2DQ2+

4DQ2+

N+

+ CH3 N

N+

N

+

CH3

N+

N

SO3-

MV2+

-O S 3

PVS

CH2 +N

N+

CH2

MV2+ PO3H2 +

N+

N

H2O3P

PAEV2+

Scheme 2.

Dye/Inorganic Nanocomposites

652 Table 5. Viologen derivatives in dye/inorganic nanocomposites [14, 18]. Methylviologen = paraquat = 1,1′ dimethyl-4,4′ -bipyridilium Diquat = N ,N ′ -ethylene-2,2′ bipyridylium = 2DQ2+ N ,N ′ -tetramethylene-2,2′ bipyridylium = 4DQ2+ Benzylviologen = 1,1′ dibenzyl-4,4′ -bipyridilium = BV2+ Propyl-viologen sulfate = PVS Ethylviologen diphosphonic acid = PAEV2+ [192] Note: Underlined references refer to intercalation in framework structures. References behind a species are additional references to those given in the headlines.

this topic (see [14] and Table 10, shown later). However, the viologen chromophor can be modified easily by the exchange of the organic residue on the pyridine nitrogen atoms. Depending on the shape and the functional groups at the end of these substituents they can form densely packed arrangements and thus be used as structure directing agents in molecular self-assembly. The potential with this respect has been demonstrated several times [69, 72]. The great potential of viologens as an electron mediator in electrochemical and photoelectrochemical processes led to trials to intercalate viologens in a great number of host lattices (see Table 10).

3.3.4. Anthracene Related Heterocycles There is a series of heterocycles which are related formally to anthracene and where the CH groups in positions 9 and 10 are replaced by heteroatoms N, O, and S. X = N are the 9 Y

X 10

acridines, X = O are the xanthenes; X = Y = N lead to the so-called phenazines, X = N, Y = O the phenoxazines, and X = N, Y = S the phenothiazines. There are a great number of compounds distinguished by substitution of various functional groups in other ring positions. All of them are strongly colored dyes, which are easily soluble in water. The compounds inserted in inorganic host lattices are summarized in Table 6; the molecular structure of some of them is shown in Scheme 3. Methylene blue (MB) is one of the most studied compounds and it has been used in clay chemistry for a rather long time [14, 32, 225–235] to investigate surface charge density and ion exchange capacity. In nearly all studies the methylene blue lay parallel to the host lattices preventing any aggregate formation in the interlayer space. Only in 2HTaS2 [132] and in clays containing residual metal ions [226] have higher layer distances been observed indicating higher packing densities and an arrangement with molecules tilted toward the layers. Photophysical studies are restricted to the investigation of methylene blue aggregation and methylene blue clay interaction (including colloidal dispersions) [14, 225–235]. (Reference [234] gives an excellent overview on the investigation of methylene blue clay interactions up to 2001.) As many other dyes of the anthracene related heterocycles, especially the phenothiazines, MB is redox active

[238–240]. Despite its electrochemical properties and its possible role as photosensitizer and/or electron acceptor it has been used in photofunctional redox chains only rarely [227, 241]. This may be related to the fact that the reduced form of methylene blue is a water insoluble neutral species which forms a conductive film on the surface of electrodes and whose behavior in electrochemical cycles is not well understood up to now [240]. Its intercalation (and also of other phenoxazin, phenothiazin, and related compounds; see Table 6) in zeolites has been studied recently [19, 149, 150, 213, 224]. Other species of this type, which have been investigated more thoroughly, are various rhodamines. Its arrangement in the interlayer space of clays and other layered material cannot be as simple as in the case of methylene blue because of the bulky phenyl substituent (see [88, 89, 206–208] and Section 4.4.2). Rhodamines are interesting pigments and are used as dyes for lasing and show the photochromic effect. These promising aspects may trigger the study of rhodamine/solid interactions.

3.3.5. Triphenylmethane Dyes Triphenylmethane dyes form a very important class of commercial dyes. They were among the first of the synthetic dyes to be developed and find applications in many fields [242]. Only a small amount of this type of dye has been brought in contact with inorganic solids relevant in this chapter (see Table 7); the most important is crystal violet (CV) (Scheme 4). Most of the studies concerned with CV/clay interactions are carried out at concentrations strongly below the cation exchange capacity (staining effects) [32, 210]. Thus, the aggregation phenomena observed with photophysical methods (metachromasy) are indicative more for adsorption than for intercalation. A series of publications is devoted to the competition of CV adsorption with respect to other dyes [202, 244, 245]. True intercalation as indicated by changes in layer distances has been observed for 2H-TaS2 [130] and -Zr–phosphate [243].

3.3.6. Azo-Compounds and Stilbenes Due to its molecular shape these species have a strong polarity, which makes them ideal components for SHG [14, 161, 162]. The presence of trans-double-bonds allows photoinduced isomerizations; the concomitant change in light absorption makes them interesting photochromic materials [14, 160]. The elongated shape of these molecules in combination with suitable functional groups at head and tail makes them ideal building blocks for the formation of mesostructures [74] and more complex supramolecular assemblies [69–72, 259, 260]. The compounds inserted in hosts or in inorganic/dye assemblies are collected in Tables 7 (charged species) and 8 (neutral species); the structural formulas of some of them are shown in Scheme 5. A very peculiar system is the donor–acceptor-substituted stilbene DAMS+ . This type of stilbene forms the largest class of organic compounds with large second-order-harmonic polarizabilities [161, 162]. The intercalation of this cation via ion exchange into CdPS3 and MnPS3 induces a spontaneous poling giving rise to an efficiency of 750 times that of urea in second-harmonic generation. The manganese derivative displays a permanent

Dye/Inorganic Nanocomposites

653

Table 6. Anthracen shaped heterocylcic dyes in dye/inorganic nanocomposites [14, 15, 19, 32]. Acridine-type dyes

Xanthene-type dyes

Proflavine Acridine orange [203, 210, 232, 236, 237] Acridine yellow

Xanthene-type dyes

pyronine Py [215] pyronineY [216] rhodamine B [204, 250] rhodamine 3B [204, 201] rhodamine 590 [208] rhodamine 640 [88, 89] rhodamine 6G [88, 89, 206, 207, 217, 218, 220, 221] rhodamine BE50 [204] rhodamine 110 [222]

fluorescein [106] fluorescein 548 [205] erythrosin B [205] eosin Y [205, 219]

oxazine 4 [88, 213] oxonine Ox+ [215] Nile blue A cresyl violet neutral species

thionine [211–213] monomethyl thionine dimethyl thionine trimetylthionine methylene blue [126, 149, 150, 202, 213, 224, 225–235] new methylene blue tetraethyl thionine toluidine blue

+

Phenazine-type dyes Safranine T [209] Janus green

Nile red [223] resorufin [214]

Note: Underlined references refer to intercalation in framework structures; italic references refer to encapsulation in gels. References behind a species are additional references to those given in the headlines.

magnetization below 40 K. Thus these materials exhibit a large optical nonlinearity and magnetic ordering [249–251]. Other stilbene derivatives (called also hemicyanine dyes) have been intercalated or included recently in Na–fluorine mica [252] and in mesoporous materials [19]. The closely related azo-benzene based dyes with aminefunctional groups have been intercalated in layered hosts like V2 O5 [127] or directly produced in the mesoporous

Et Et

H H

N

H

H

N

+

N

O

N

H

Et

Et

N

H

+

N

O

Et

CO2Et

Rhodamine 3B

Rhodamine B

H Et

CH3

CH3

CO2Et

H N H

N

+

N

H

CH3

N

CH3 +

N

N

N H

O

CH3

+

CH3

H

N

N H + N H

N

N

Acridine orange

H

N

H CH3

H

H

CH3 N H CH 3 H

Proflavine

Rhodamine 6G

H

Et

CO2H

Pyronine

Et

+

O

H

+

O

N

N

+

O

N

N

H H

CH3

Oxonine

Nile blue A

Safranin T N

N

CH3 N

S

CH3

Methylene blue

Scheme 3.

N

+

CH3 H N CH3 H

S

Thionine

N CH3 H N N H CH 3 +

S

Trimetyl thionine

N

+

H CH3

Y-zeolitebs [254]. The unsubstituted azo-benzene has been inserted in AlPO4 -5 and ZSM-5 [255]. The need for better organization of organic chromophores in layered inorganic/organic composites leads to the intercalation of single chain ammonium amphiphiles having an azo-benzene chromophor in the hydrophobic chain [93, 256–258].

3.3.7. Other Types of Dyes There are numerous other dyes which have been inserted in solid hosts. Because they belong to various classes of organic compounds and they are not studied in depth they are mentioned only for the sake of completeness and are collected in Tables 7 and 8. The structural formulas of the more interesting species are given in Scheme 6. In the following we will give only a few comments on special examples or groups of compounds. Whereas cyanine dyes play an essential role as photoactive components in Langmuir–Blodgett assemblies [60–63, 65, 66], intercalation of cyanine dyes in layered compounds have been rare up to now [264, 265]. The extent of intercalation of aromatic ammonium compounds (e.g. naphthyl, pyrenyl ammonium) is rather low because of their hydrophobicity [14, 15, 263]. If, however, the solid is preloaded with a surfactant system the extent of adsorption increases dramatically [14, 15, 263]. Using the photophysical properties of the aromatic amines it is possible to probe their aggregation behavior and the surface properties of the inorganic support. It is not necessary to use the aromatic amines; one can also use the corresponding aromatic hydrocarbons instead if the surface is made hydrophobic by preadsorption of surfactant species. One of the most interesting probe molecules is pyrene due to the high quantum yield of fluorescence, excimer-forming capacity, and great sensitivity of its photophysical behavior to its

Dye/Inorganic Nanocomposites

654 Table 7. Positively and negatively charged dye species in dye/inorganic nanocomposites. Triphenylmethane derivatives [32, 14, 15] CPh3 [130] Michler’s hydrol blue [247]

+

fuchsine crystal violet [210, 243, 247, 248] methylgreen [245, 247]

pararoseaniline ethyl violet malachite green [210, 246, 248] brilliant green [210]

Stilbenes or hemicyanins DAMS [249–251] = 4DASPI [97] = trans-4-[4-(dimethyl-amino)styryl]-1-methylpyridinium 2DASPI [252] = 2-[4-(dimethyl-amino)styryl]-1-methylpyridinium LD698 [252] = 1-ethyl-2-(4-(p-dimethylamino)phenyl-1,3-butadienyl)pyridinium LD722 [252] = 1-ethyl-4-(4-(p-dimethylamino)phenyl-1,3-butadienyl)pyridinium Stilbene 32 [19] Stilbene 33 [19] Stilbene 34 [19] Azo-compounds Methyl yellow [127] Methyl-orange [105] 4-[N -ethyl-N -(2-hydroxyethyl)]amino-4′ -nitroazobenzene [253] C12 AzoC5 N+ = p-(-trimethylammoniopentyloxy)-p′ -(dodecyloxy)azobenzene [93, 256–258] C8 AzoC10 N+ = p-(-trimethylammoniodecyloxy)-p′ -(octyloxy)azobenzene [257, 258] Cn AzoCm N+ [74] Chrysoidine = dimethylaminoazobenzene [254] Basic blue 159 [19] Basic red 18:1 [19] 4-(2-Pyridylazo)resorcinol [261] 4-{4-[N ,N -bis(2-hydroxyethyl)amino]phenylazo}benzylphosphonic acid [259, 260] (4-{(4-(4-(4-(2-hydroxyethyl)sulfonyl)phenylazo}phenyl)piperazinyl)phenyl)phosphonic acid [260] (3-{(4-((4-(2-cyano-(2-hydroxyethyl)carbonyl)ethenyl)phenylazo}phenyl)N -methylamino)propyl)phosphonic acid [260] New coccine = Na3 -7-hydroxy-8-[(4-sulph-1-naphthalenyl)azo]-1,3-naphthalenedisulphonate [262] Arylammonium systems [15] C10 H7 CH2 NH+3 [263] C14 H9 CH2 NH+3 [263] C16 H9 NMe+3 [263] C16 H9 (CH2 )n NMe+3 , with n = 3, 4, 8 C16 H9 (CH2 )4 NH+3 [263] PBITS [75] = N ,N ′ -di(phenyl-3,5-disulphonicacid)-perylene-3,4:9,10-tetracarboxydiimide Various compounds Cyanin [264, 265] 2,4,6-Triphenylpyrylium [19, 266] Thioflavine T [202, 245] Coumarin [267] Basic yellow 40 (coumarin 40) [19] Indigo carmine [262] Anthraquinone sulfonate [272] 8-Hydroxy-5,7-dinitro-2-naphtalene-sulfonicacid [273] 8-Hydroxy-quninoline [274] Note: Underlined references refer to intercalation in framework structures; italic references refer to encapsulation in gels. References behind a species are additional references to those given in the headlines.

surroundings (see [14] and a series of original publications [290–295]). Many of the phenomena observed will happen at outer crystalline surfaces and not in the interlayer galleries. This seems to be different in case of the solid-state uptake of pyrene in intercalation compounds of long-chain alkylammonium ions [296]. In this context it should be mentioned that dye assembly [212] and formation of charge transfer complexes (between hydrophilic various pyridinium containing ions and hydrocarbons) occur also in supercages of some framework structures [10, 297]. For these aggregation

phenomena the influence of residual water has to be taken into account [212]. A very interesting class of molecules is the neutral disubstituted benzenes with the functional groups in para-position (Table 8). These species are highly polar and show in the solid state or in suitable matrices the effect of second harmonic generation. Intercalated in hydrophobic (neutral) framework structures with parallel running channels they are highly oriented [8, 268–271]. If in addition the structures are not centrosymmetric the SHG effect can be strongly amplified.

Dye/Inorganic Nanocomposites

655 CH3 N

CH3

N

+

CH3

CH3 N CH3

CH3 N

N CH3 H C +N 45 22

DAMS+

N

+

CH3

Stilbene 33

LD698 C 2H 5

CH3

CH3

S

CH3

N

N

N +

CH3

CH3

+ N

N

CH3

CH3 CH3

Brilliant green

N

+ N

CH3

CH3

CH3

N

CH3

+ N

C12AzoC5N+

CH3 N

+ N

N N

CH3

N

N

N N

Basic Red

+

OH

+

P

OH OH 4-[4-[N,N-bis O (2-hydroxyethyl)amino] phenylazo]benzylphosphonic acid

+

N N

S

N

N

N

Basic Blue 159

CH3

CH3

OH CH3

Malachite green

NH2

Chrysoidine

CH3

N

N

CH3 CH3 N

N

CH3

C5H10 N

CH3

N

Methyl green

N N

Scheme 4.

N

CH3

H N

O O

N

NaO3S

N

Methyl yellow

-

S O

N

N

Methyl orange

NaO3S H

New coccine

Scheme 5.

Table 8. Neutral systems inserted in inorganic matrices. Small bipolar donor acceptor substituted organic chromophores [8, 19, 268] p-Nitroaniline [269] N -methyl-p-nitroaniline p-Nitrodimethylaniline 2-Methyl-4-nitroaniline Phenol dyes

p-dimethylaminobenzonitrile [270, 271] 2-amino-4-nitropyridine 2-amino-4-nitropyrimidine p-nitropyridine-N -oxide Aromatic imids

Phenol red [275] Cresol red [275] Bromophenol [275]

N -methylphtalimide [246] perylene red [220] perylene orange [220]

Azocompounds trans-Azobenzene [255]

diethylazodicarboxylate [280] diisopropylazodicarboxylate [280] dibenzylazodicarboxylate [280]

Heterocyclic compounds Indigo [4–6, 159] Thioindigo [283]

OH

NH2

CH3 +

O

N

Cl

CH3

Stilbene 34 OH

N

O2N

N

CH3

+

Stilbene 32

C12H25 O

Crystal violet

CH3

CH3

N CH3

CH3

2H-spiro[6]nitrochromene-2,2′ -[1,3,3]trimethylindoline [276, 277, 278–280] 1,3-dihydro-1,3,3-trimethylspiro[2H ]-indole-2,3′ -[3H ]naphth[2,1-b][1,4]oxazine (SO) [277] other spiropyranes [277, 278, 280] 1,4-diketo-2,5-dimethyl-3,6-diphenylpyrrolo[3,4-c]pyrrole; R = H, CH3 [281]

Strong electron donors; cation forming species [125, 128, 129, 284–289] TTF = tetrathiafulvalene Tetramethyl-TTF TTT = tetrathiatetracene TTN = tetrathianaphthalene ET = bis(ethylenedithia)tetrathiafulvalene DBTTF Note: Underlined references refer to intercalation in framework structures; italic references refer to encapsulation in gels. References behind a species are additional references to those given in the headlines.

SO3Na

Dye/Inorganic Nanocomposites

656 (CH2)n +N

CH3 CH3 CH3

SO3





SO3

O

O

N

N

SO3

O

O



SO3

PBITS

CH3

S

S Et

N Et

O

O

Coumarin-1

+N

N

Et

Et

Cyanin

TTF

H N

N H

O

O

Indigo

CH3 CH3

CH3 CH3

O

O

N H

SO3Na

Indigo carmine

N

S

S

O

H N

NaO3S



NO2

N

O

R3

R1

CH3

Spirochromene

R2

Scheme 6.

The spiropyranes are the most widely studied photochromic dyes [298]. Photochromic materials have potential for optical switching and information storage. The immobilization in a clay/didodeceyldimethylammonium matrix was tried as early as 1991 [278]. Recently these interesting species have been encapsulated in inorganic and organic matrices [276, 277]. Very interesting is the intercalation of electron donors including TTF, the most famous donor for preparing organic conductors. However, the intercalation of these species is restricted to silicates [284], FeOCl [128, 129], V2 O5 [125], and some MPS3 compounds [285, 286], hosts which show a redox potential positive enough to accept electrons from these donors or do not lead to a reduction of the cation intercalated. The proposed intercalation of DBTTF into layered dichalcogenides [287] is hard to evaluate with respect to this condition. In case of Zr–P surfactant long-chain alcohols have been used to expand the host structure allowing the intercalation of partially charged TTF [289]. Just the cations can be considered as dyes. There are no further investigations of the photophysical properties of these intercalation compounds.

4. PREPARATION OF DYE–INORGANIC HOST COMPOSITES The methods of preparation of dye intercalation compounds are the same as used in the intercalation processes of inorganic ions and small organic molecules [19–21, 78–83]: • Adsorption of organic neutral compounds: This can be achieved by treating the solid with the species to be inserted in the gas, the liquid, and the solid state, or out of a solution. It should be carried out at moderate temperature because the species adsorbed is in equilibrium

with the surrounding. If the species intercalated is very volatile some precautions against outgassing have to be undertaken. Solid compounds with low vapor pressure can be inserted out of solutions or by solid-state reactions as shown for the intercalation of pyrene in clays preloaded with surfactants [296]. The species of interest in this chapter can be taken up only in host lattices with a hydrophobic surface, which means it is restricted to special frameworks of the AlPO4 family [8, 19, 268– 271]. In the latter case the uptake of volatile dyes is restricted to those for which the smallest diameter is smaller than and or of the same size as the effective window size of the channels. • Ion exchange [19–21, 79, 83]: This is the most convenient procedure for intercalation of the cationic and anionic dyes and is carried out by soaking the solid with an aqueous solution of the dyes. The reaction is carried out normally at room temperature. Since many of the dye ions are bulky the uptake in the interlayer space is slow unless there is a strong interaction of the electron system with the surface of the host lattices. Due to this hindered uptake in the interlayer space the adsorption on external surfaces competes with it (especially if the particles are in the m range or the interaction with the organic species is very strong). This method can be applied for the intercalation of nearly all dye ions in many of the charged layered host compounds and also for zeolites or zeolite related materials as long as the frameworks have an excess negative charge. However, in the latter case the uptake is again restricted to those dyes for which the smallest diameter is smaller than or of the same size as the effective window size of the channels. • Adsorption on dispersed solids: This is a very reasonable method for all layered host systems which form colloidal dispersions [83, 90–92]. In the ideal case all layers are separated from each other and form monolayer dispersions. In this case the surfaces of all particles can be considered as external, at least below the critical flocculation concentration (often identical with the cation exchange capacity = cec). This state is ideally suited to investigate the dye/solid and the dye/dye aggregation phenomena by visible light [7, 9, 14, 15, 290–296]. In real systems not all layers will be separated from each other, but there will be distribution of stacks with different numbers of individual layers (Fig. 13). Thus the uptake in the interlayer space competes with the adsorption on external surfaces. Since the dye molecules interact very strongly with the surfaces of the inorganic hosts, at concentration higher than the cec dye species will be found on the external surface of the solid particles, in the interlayer space, and in voids of tactoids formed during flocculation. This method is ideally suited for the preparation of inorganic/organic composites with very bulky dye molecules, if the orientation of the species is not of importance. • “Ship in the bottle” synthesis [18, 19]: The construction of molecules in the zeolite itself is a further strategy to overcome the problem of the finite size of the accessible window. With this approach it is possible

Dye/Inorganic Nanocomposites Elementary sheet parallel view

657

O N

= Elementary sheet top view

O N

Interlayer space

O

N

O N N

O N

O

Regular stack

O N

N Swollen regular stack

N N Zn N

Aqueous colloid O

N N N

N

NO

N O N

Turbostratic stack ON

O

N

NO

Mesopores

Tactoid or film

Aqueous gel Aggregate or dry powder

Figure 13. Sketch of the various forms of clay dispersions and their flocculation products.

to get bulky species in the interior of the solid. It is applied for example to the synthesis of porphyrins in clay minerals [170], of polypyridine complexes of transition metals [18, 19], the synthesis of a spiran molecule within synthetic faujasites [276], the azo coupling in zeolite Y [254], the synthesis of Nile red in zeolites [223], and the formation of the triphenylpyrylium cation (an interesting electron-transfer sensitizer) in zeolite Y [266]. The species produced are so bulky that they cannot escape out of the solid; thus these compounds should be considered inclusion compounds and not intercalation compounds [83]. • Dye species as templates: A breakthrough for the insertion of dye species in framework structures is the insitu inclusion of these organic complexes during synthesis of the frameworks [19]. Thus, methylene blue [19, 224] or the cobaltocinium complex [136, 163] has been used as template. In case of the methylene blue one methyl group is missing after synthesis. To avoid the decomposition of sensitive dyes in the highly aggressive hydrothermal media a change of gel composition and the assistance of microwaves (reducing the reaction time and temperature) is helpful [19]. Another approach is the modification of the dyes with such substituents that they can be included in the template micelles. This method has been applied for the synthesis porphyrin-loaded mesoporous solids [153]. Figure 14 shows a sketch of the phthalocyanin arrangement in the hexadecyltrimethylammonium chloride micelles. The phthalocyanins have also be used as templates in the synthesis of new mesoporous solids based on some transition metal oxides [154–156]. In recent publications chromophores have been substituted with long tails of alkyl chains with polar head groups. These new classes of dyes are amphiphilic and act as structure forming templates in molecular sieve synthesis [73–75]. It should be mentioned that this technique was applied for the first time in 1991 to form an intercalation compound of porphyrin in a synthetic clay [184]. • In case of the electronically conducting 2H-TaS2 the intercalation can be carried out electrochemically [83, 131]. A complete reaction can be achieved under

Figure 14. Sketch of the arrangement of a modified Zn–phthalocyanine in the hexadecyltrimethylammonium micelles used in the synthesis of mesoporous Si-MCM-41 and Ti-MCM-41 frameworks. Adapted from [153].

galvanostatic conditions overnight at room temperature. Normally the reaction is carried out with 2H-TaS2 crystals of a few millimeters in diameter. This gives intercalated samples with a residual resistance ratio in the order of 10 to 40. Depending on the intercalation current a different packing density of the dye molecules can be realized, at least for the case of methylene blue intercalation. In almost all these cases one gets powder materials without preferred orientation. For many applications highly ordered macroscopic assemblies are desired. The following methods have been developed in order to get macroscopic assemblies: • Oriented aggregate technique: Colloidal dispersions can be used to deposit thin films of solids on suitable substrates (glass, ITO coated glass, metal electrodes, etc.) by evaporating to dryness a 2% suspension. Such oriented supported or self-supporting films have been applied in clay chemistry for many years [83, 87, 88, 172, 173, 201, 209, 267, 299, 300]. Such films can be loaded easily with desired species. However, there is again the problem of competition of external adsorption and deposition in mesoporous voids and in the interlayer space as in case of the colloid preparation mode (for structure see Fig. 13). A very recent application of such dye-loaded films is the preparation of clay-modified electrodes [55, 56, 196]. • layer-by-layer deposition [72, 301]: The idea of preparing electrostatically bound thin films was introduced by Iler in 1966 [302]. He showed that surfaces modified with positively charged boehmite colloids or cationic amphiphiles could bind negatively charged colloidal silica or latex particles (for a schematic view of such a deposit see Fig. 15). This method was applied later for sequential adsorption of alternatively charged polyelectrolytes [301, 303, 304]. It can be adapted to the sequential layering of inorganic polyanions (obtained by exfoliation of layered intercalation compounds: clays, ternary oxides, Zr-phosphates [92]) with a variety of polymeric cations [72, 304–310]. The first inorganic polyanionic layer has to be deposited on a self-assembled organic bifunctional layer anchored by covalent bonding to a solid substrate as shown in

Dye/Inorganic Nanocomposites

658 OCH3

(a)

A

H2N(CH2)4Si CH 3 CH3

B C

or H2N(CH2)2SH

D E

substrate (Si, Au, ...)

F

 exfoliated α-ZrP, Ti2NbO7–, etc.

G

multilayer thin film

(b) DM DP/M

DP

Scheme 7. 2 nm 100 nm

(c)

3.6 nm

1.0 nm

1.0 nm 5.5 nm 4.6 nm

substrate

Figure 15. Layer by layer deposits. Schematic views of various types of deposits: (a) The classic schematic cross-section of a multilayer film; A, C, and E represent layers of 100 m silica; B, D, and F represent layers of colloidal boehmite fibers; G the glass substrate. Adapted from [302]. (b) An oversimplified representation of a polymer/montmorillonite layer: DP/M is the mean repeating distance of a P/M layer, DM is the mean thickness of a montmorillonite layer, and DP is the mean thickness of a polymer layer. Adapted from [309]. (c) Schematic drawing of a hybrid film consisting of a synthetic lithium hectorite and a Ru/bipyridine complex, carrying long aliphatic sidegroups at the bipyridine ligand. Adapted from [319].

Scheme 7. On this assembly the first layer of a polymeric cation can be adsorbed followed by the second inorganic sheet. In repeating adsorption/washing procedures subsequent layers of alternative charge can be deposited. However, the real multilayer deposits will not be as regular as expected. (a) Again the real colloids will not be so monodisperse that only monolayers are present in the solutions. (b) It will be impossible to tune the deposition in such a manner that only a regular monolayer of the disperse system will be formed. A tactoidlike arrangement is more probable. (c) An interdigiting of the polymer and the colloid particles can also not be excluded. Just these sorts of defects have been observed recently [309, 310], and the term “fuzzy assembly” has been introduced for this sort of multilayer architectures. For the purpose of this chapter it is important to note that it was possible to form a clay/polymer/dye film by this method recently [311].

In case of the very peculiar Zr–phosphate phosphonate system it could be shown that the technique discussed before can be used for the self-assembly of low molecular species also [71, 72]. This is based on the following experimental facts: during precipitation of the Zr–phosphate the phosphoric acid can be replaced with phosphonic to a considerable amount without severe changes of the inorganic backbone formed; the affinity of the zirconium metal ions to the phosphate groups is exceptionally high and, thus, the solubility of the building blocks extremely low. From structural analogy of Zr–phosphonates with aliphatic chains and Langmuir–Blodgett films it was concluded that it should be possible to prepare multilayer assemblies on a solid substrate if it is possible to form stable assemblies of phosphonic acids on a solid substrate. Cao and co-workers demonstrated this in 1988, for the first time [71]. Since then many multilayer assemblies of this Zr–phosphonate system have been obtained including many with chromophores in the organic backbone of the phosphonic acid. Typical chromophores are azobenzene derivatives and viologens (typical examples are shown in Fig. 16). A very interesting multilayer assembly has been prepared with a viologen spacer [69, 314]: it contains halide ions compensating the positive charges of the pyridinium nitrogen atoms. If this sample is irradiated in vacuum or under argon the viologen is reduced immediately as indicated by the blue color of the viologen radical. • Langmuir–Blodgett technique: This method is based on the self-assembly tendencies of membrane building amphiphilic molecules. In 1993 Lvov et al. combined this method with the polycation/polyanion based layer-by-layer deposition. Polyelectrolytes form stable complexes by electrostatic interactions with some amphiphiles, which spread to a homogeneous film at the water–air interface. These complexes can be transferred by the Langmuir–Blodgett (LB) technique to solid substrates [315]. In a very similar manner clay particles intercalated and embedded with bulky amphiphilic cations. The clay can be spread (out of a solution in volatile organic solvent like chloroform)

Dye/Inorganic Nanocomposites

HO

a Zr

659

4+

+ N X-

+ (HO)2(O)P

+ N X

O N

P(O)(OH)2

Dye monolayer

N N

HF/H2O

Zr O3P X

+ N

+ N

XPO3

b P

X-

P

X-

P

X-

P

X-

O Zr O

O O P O HO

P O O Zr O

Zr phosphatephosphonate interlayer

O P O O

N

OH

N

P X- P X-P X-P X-

Dye monolayer N

N+ N+ N+ N+

N+ N+ N+ N+ X- P X- P X- P X- P P

X-

P - P -P X X X

O O P

O Zr O

N

N

N

O P O O O P

Zr O

O O P O

Zr phosphatephosphonate interlayer

SiO2

Figure 16. Schemtaic structural view of Zr–phosphonate multilayers. Left: Layer-by-layer deposit made from substituted viologens and Zr ions; by illumination a stable charge separation can be produced. Right: Layer-by layer deposit of an Zr–diphosphonate containing azobenzene groups; this system exhibits SHG. Adapted from [312].

on the water surface in an LB trough. This spread clay film can be transferred on hydrophobic glass or carbon [316, 317]. Even alkyl ammonium cations with short alkyl chains (C4 –C12  form hybrid monolayers with clay particles at the air–water interface [318]. In the meantime it was also possible to spread a clay/Ru(phen)2 (dC18 )bpy complex at the air–water interface to form a multilayer film of this complex by the LB technique [319]. • Orientation and oriented growth of framework crystals: For applications of dye/inorganic complexes in framework structures it is not enough to have the photosensible organic dyes aligned in parallel channels of frameworks, but it is also necessary to have aligned the whole crystals. Various methods for aligning crystals of zeolites have been discovered since about 1990 [320]. The materials obtained were of interest for catalysis, sensor applications, and separation technologies. We would like to mention only two examples, which are of interest in the context of this chapter: the crystal growth of AlPO4 on an organophosphate film [76, 321] and the alignment of ZSM-5 and AlPO4 -5 in an applied electric field [322]. After aligning the crystals are fixed by water glass. Such oriented crystals loaded with p-nitro aniline show a SHG intensity of 100%, which is a factor of 5 higher than in a random powder assembly [322].

5. STRUCTURE AND STRUCTURE SIMULATION The knowledge of structure and understanding of the structure–property relationship is crucial for the design of new intercalation compounds with desired properties. These properties can be tuned by the proper choice of the host–guest combination, the guest concentration, and cointercalation of further guest species. However, the structure analysis of layered intercalates brings specific problems. As the intercalation is in fact an insertion of a known molecule

into a known layered crystal structure, we have to determine the positions, orientations, and arrangement of the guest molecules in the interlayer space, the way of layer stacking, and possible changes in conformation of guest molecules. Intercalated structures exhibit very often various degrees of disorder in arrangement of guest molecules and in layer stacking. Consequently the characterization of the disorder usually belongs to the structure analysis of intercalates. The crystal packing of intercalated structures is a result of the host–guest complementarity [323–325], which reflects the character of the host–guest interaction and the geometry of the host structures and guest species in the pristine state. Factors governing the crystal packing in intercalates can be summarized as follows. • the nature of the active sites on the host structure and guest species (i.e., the charge distribution on the host structures and guest species, the type of functional groups, etc.) • the ratio between the host–guest and guest–guest interaction energy • the topology of host structure (i.e., the distances and ordering of functional groups, the presence of steric barriers for the diffusion of guest molecules, etc.) • the size and shape of the guest molecules, especially the ratio between the size of guests and distances of the active sites in the host structure • the rigidity (or flexibility) of host structure and guest species (i.e., the tendency to distortions in the host structure and to conformational changes of guest molecules due to intercalation) These factors act sometimes in concert and sometimes in conflict. As a result of this interplay we can get a perfectly ordered intercalated structure in case of perfect host–guest complementarity. The requirement of structure ordering is of great importance in a design of new intercalates for special applications, in which one has to control the porosity in intercalated layered structures (selective sorbents, molecular sieves), or to control photofunctions of guest molecules, etc. On the other hand, the departure from the perfect host– guest complementarity leads to various degrees of disorder in an intercalated structure.

5.1. Method and Strategy of Structure Analysis Intercalated layered structures exhibit very often various degrees of disorder in positions and orientation of guest molecules and consequently in layer stacking. This disorder obstructs the structure analysis based only on the diffraction method. Intercalation of organic guests into an inorganic host structure introduces an additional specific problem into the diffraction analysis of intercalates. Scattering amplitudes of carbon, oxygen, and hydrogen are small in comparison with the atoms building the inorganic host structures and consequently the contribution of guest molecules to the total intensity diffracted from the crystal is too small. This fact complicates the precious localization of guests in the intercalated compounds. For disordered guest molecules, which in addition contribute very little to the diffracted intensities, the diffraction analysis fails. In such a case the molecular modeling

660 (molecular mechanics and molecular dynamics) represents a very powerful tool in structure analysis, provided that the modeling is combined with an experiment. Experiment plays an important role in creation of the modeling strategy and in corroboration of modeling results. In present work we combine the modeling in Cerius2 [326] with X-ray powder diffraction and infrared (IR) spectroscopy. Using the combination of structure simulation and experiment can provide us with a detailed structure model including the characterization of the disorder. In addition to diffraction analysis we get the total crystal energy, the total sublimation energy, and the host–guest and guest–guest interaction energy.

5.2. Modeling of Intercalates Molecular modeling (molecular mechanics) is a method of optimization of the structure and bonding geometry using minimization of the total potential energy of the crystal or molecular system. The energy of the system in molecular mechanics is described by an empirical force field [327, 328]. This simplified description of the crystal energy enables us to model large supramolecular systems, which cannot be treated using ab initio calculations. The energy terms describing the valence interactions (i.e., the bond stretching, angle bending, dihedral torsion, and inversion) may differ in various force fields [326] and it is evident that the choice and test of the force field belong to the most important part of the modeling strategy. Two methods are used to test a force field. First of all the comparison of force field calculations with experiment is the most reliable test of the force field. For this purpose one can use the known structure from the database, which is similar to the investigated structure. In case of intercalation compounds it is difficult to find related structures in databases and then we have to use the limited number of structure parameters obtained from the X-ray powder diffraction pattern affected by a disorder. The second method of force field testing is comparison with the results of ab initio calculations for a small fragment of structure [329]. The force field library in Cerius2 [326] contains the universal force field [330] and several special force fields designed for modeling of organic supramolecular systems [331, 332], polymers [333], and sorption of organic molecules into silicates and zeolites [334]. In the Cerius2 modeling environment the atomic charges are calculated using the Qeq (charge equilibration) method [335] and the Ewald summation method [336, 337] is used to calculate the Coulomb energy. The Lennard-Jones potential is used for calculation of van der Waals interactions. Classical molecular dynamics calculates the dynamic trajectory of the system solving the classical equations of motion for a system of interacting atoms [326–328]. The temperature and the distribution of atomic velocities in the system are related through the Maxwell–Boltzmann equation. Introducing the temperature and pressure controlled molecular dynamics enables one to study temperature dependent processes and to explore the local conformational space. However, for crossing large energy barriers, molecular dynamics at high temperatures can be used. All the computational methods searching for the global energy minimum have to generate the large number of initial

Dye/Inorganic Nanocomposites

models using three different methods [327, 338]: • deterministic method for generation of starting models, performing the systematic grid search that covers all areas of the potential energy surface [338–340] • molecular dynamics generating the starting geometry [326–328] • stochastic methods (Monte Carlo [328], genetic algorithm [341]) The first method, performing the systematic grid search, can be used easily to generate starting models in case of small or rigid large guests and rigid host layers. On the other hand molecular dynamics is very well suited for generation of starting models in case of large flexible organic guest molecules. For modeling of layered intercalates it is useful to combine the systematic grid search with the molecular dynamics to generate the staring models. The strategy of modeling consists of: the building of initial models, the setup energy expression, the choice and test of the force field, the definition of rigid fragments, and the setup of fixed and variable structure parameters, etc. Reliable results of modeling can be obtained providing that all the steps in the modeling strategy are based on available experimental data.

5.3. X-Ray Diffraction and Vibration Spectroscopy Because of the disorder in intercalated structures, it is usually impossible to prepare a single crystal of reasonable size suitable for X-ray diffraction measurements. The powder diffraction pattern, which is influenced by disorder and a small scattering contribution of guest molecules, is in addition affected by the sample effects like preferred orientation of crystallites [342] and absorption in surface roughness [343]. In spite of these difficulties, the diffraction pattern can always provide us with some information useful for the strategy of modeling. At least we can get the basal spacing (interlayer distance), indicate the disorder, and recognize the type of disorder. In more favorable cases we can even get the lattice parameters of the intercalated structure. These experimental data are useful for the building of the initial models. Comparing the diffraction pattern for the intercalated and host structures one can deduce the changes of structure after intercalation and consequently to set up the most suitable modeling strategy (variable and fixed structure parameters, rigid fragments, rigidity of layers, etc.). The diffraction line profiles indicate a possible lattice strain due to the disorder in arrangement of guests and the disorder in layer stacking. Comparing the vibration (IR/Raman) spectra of the intercalate, the pristine guest compound, and the host compound we can see possible changes in the bonding geometry and the conformation of guest molecules and host layers, which may occur during the intercalation. If the spectral bands corresponding to the host layers exhibit the same positions and profiles in the spectrum of host structure and intercalate, one can conclude that there are no changes in bonding geometry of host layers during intercalation. This is a very

Dye/Inorganic Nanocomposites

important conclusion for the modeling strategy, as in such a case the host layers can be treated as rigid units during energy minimization. The same conclusion can be derived for the guest molecules. In case of rigid host layers and rigid guests we obtain the IR spectrum for this intercalate as a superposition of the host structure and guest compound spectra. The rigidity of host layers is the crucial assumption in modeling of layered intercalates, as in case of inorganic covalently bonded host layers the force field calculations cannot be used to reveal possible changes in their bonding geometry.

5.4. Examples of Structure Analysis The perfect three-dimensional (3D) order in intercalated structure requires the perfect ordering of guest molecules in the interlayer space and consequently the perfect order in the layer stacking. The ordering of guests requires the regular network of active sites on the host layer which means the presence of functional groups or regularity in charge distribution on the host layer. This effect will be illustrated on three examples: tantalum sulphide intercalated with methylene blue, monmtorllonite intercalated with rhodamine B, and vermiculite intercalated with anilinium.

661 of MB cations (for MB cations see Fig. 17a). The concentration of the MB cations in the initial models was set up according to the chemical analysis, performed in previous work [346], which means: • For the phase I-MB-TaS2 there are two MB cations per one two-layer 3a × 5b × 1c supercell. • For the phase II-MB-TaS2 there are four MB cations per one two-layer 3a × 5b × 1c supercell. • For the phase III-MB-TaS2 there are six MB cations per one two-layer 3a × 5b × 1c supercell.

A systematic grid search combined with molecular dynamics was used to generate the series of initial models for energy minimization in Cerius2 . The energy minimization of all initial models was carried out in the space group P 1 under the following conditions: the host layers were treated

(a)

5.4.1. Tantalum Sulphide Intercalated with Methylene Blue Tantalum sulphide 2H-TaS2 has been found as a suitable host lattice for intercalation of organic or organometallic molecules (for reviews see [78–83]). The discovery of superconductivity in the intercalation compound of 2H-TaS2 initiated a very broad investigation of these compounds [344, 345]. Recently, it was shown that electrointercalation of MB [131, 346] into 2H-TaS2 leads to three different phases indicated by different basal spacings and superconducting transition temperatures Tc . (The intercalation compounds will be denoted in the forthcoming text as MB-TaS2 .) The first phase, I-MB-TaS2 , exhibits an interlayer distance 9.07 Å, obtained from X-ray powder diffraction (XRPD) measurements. Since this phase shows the two-layer stacking as the host structure 2H-TaS2 , the parameter periodicity d001 in the direction perpendicular to the layers is 2 × 9 07 C = 18 14 Å. The Tc for this phase is 5.21 K (slow cooling) and/or 4.47 K (fast cooling). (For more details see [131, 346].) For the second phase, II-MB-TaS2 , the interlayer distance obtained from XRPD measurements is 11.9 Å and the Tc for this phase is 4.92 K. For the third phase, III-MB-TaS2 , the interlayer distance obtained from XRPD measurements is 16.80 Å. The Tc for this phase is 4.24 K. From experience with 2H-TaS2 intercalation compounds we conclude that phase II and phase III should also have a two-layer periodicity in the direction perpendicular to the layers. The initial models of MB-TaS2 were built using the known structure data for the host compound 2H-TaS2 (hexagonal, space group P 63 /mmc with Ta in trigonal-prismatic coordination by S) [347]. The unit cell (cell parameters: a = 3 314 Å and c = 12 097 Å, and Z = 2, i.e., two formula units TaS2 per unit cell) contains two TaS2 layers mutually shifted. The supercell 3a × 5b × 1c was built for the intercalation

(b)

Figure 17. (a) Structure of tantalum sulfide intercalated with methylene blue phase I-MB-TaS2 , side view of one 3a × 5b × 1C supercell. (b) Top view of structure of the phase I-MB-TaS2 , illustrating the position of the MB cation on the TaS2 host layer and the layer stacking.

662

Dye/Inorganic Nanocomposites

as rigid bodies, but they were allowed to shift with respect to each other. The cell parameters a$ b and angle % were kept fixed. Variable parameters were the supercell parameters c, , &, and the bonding geometry, position, and orientation of the MB cations were also variable. (For more details see [132].) The structure of the phase I-MB-TaS2 obtained as a result of molecular modeling is shown in Figure 17a,b. This model represents the global minimum of the crystal energy. No multiple minima have been detected; that means the structure exhibits the perfect 3D ordering. As one can see from the figure, the planar MB cation resides just in the central plane in the interlayer space, perfectly parallel with the host layers. The calculated interlayer distance d = 8 99 Å is in good agreement with the experimental value 9.07 Å. The calculated supercell parameters are A = 10 02 Å, B = 16 70 Å, C = 17 98 Å,  = 90 0 , & = 90 0 , and % = 60 . Figure 17b shows in detail the position of the MB cation with respect to the lower and upper host layers. This position is a result of the predominant electrostatic host–guest interaction in this intercalate, where the negatively charged atoms in the MB cation tend to avoid the negatively charged sulfur atoms in the host layer. Figure 17b illustrates in addition the layer stacking in the phase I-MB-TaS2 . Figure 18 shows the structure of the phase II-MB-TaS2 , the side view of the two-layer 3a × 5b × 1c supercell with four MB cations obtained as a result of molecular

simulation. The calculated supercell parameters are A = 10 05 Å, B = 16 75 Å, C = 25 36 Å,  = 90 0 , & = 90 0 , and % = 60 ; this means the interlayer distance is 12.68 Å. The MB cations are tilted with respect to the host layer creating pairs of discrete cations [C16 H18 N3 S]+ in antiparallel arrangement. This pairlike arrangement of MB cations with the same relative positions and orientations occurs also in the crystal structure of MB-thiocyanate [348]. According to the results of modeling the structure of II-MB-TaS2 is ordered. Energy minimization for the series of initial models led always to the same perfectly ordered layer stacking and to the same orientation and position of MB cations with respect to the TaS2 layers. As one can see in Figure 18, the longer edge of the planar MB cation resides in the groove between the rows of sulfur atoms. This arrangement of guests results in a layer stacking different from phase IMB-TaS2 , as one can see comparing Figures 17 and 18. (For details see [132].) Comparing the values of the sublimation energy per supercell, one can see that phase II-MB-TaS2 is significantly less stable than phase I-MB-TaS2 . This result of modeling is in agreement with the experimental results [346]. The arrangement of the MB cations in phase III-MBTaS2 is illustrated in Figure 19. Energy minimization for the series of initial models led always to same layer stacking and to the same orientation and position of MB cations with respect to the TaS2 layers. That means that the structure of this phase is ordered. The calculated supercell parameters are A = 10 05 Å, B = 16 75 Å, C = 33 33 Å,  = 90 52 , & = 93 48 , and % = 60 . The calculated interlayer distance of 16.63 Å is in good agreement with the experimental value of 16.8 Å. Phase III-MB-TaS2 keeps the same character of layer stacking, as phase II-MB-TaS2 . The supercell in this case is not rectangular ( = 90 52 , & = 93 48 ) and consequently two successive layers are mutually slightly shifted from the ideal arrangement in Figure 17. The total sublimation energy for phase II-MB-TaS2 is nearly the same as in case of phase III-MB-TaS2 .

Figure 18. Structure of tantalum sulfide intercalated with methylene blue phase II-MB-TaS2 .

Figure 19. Structure of tantalum sulfide intercalated with methylene blue phase III-MB-TaS2 .

Dye/Inorganic Nanocomposites

5.4.2. Montmorillonite Intercalated with Rhodamine B The intercalation of xanthene dyes in the smectite-type clays is followed by the changes observed in the absorption spectrum (the so-called metachromic effect), whereby the main absorption band is shifted to higher energies [206, 207, 216]. This effect is explained by: (i) the interaction between the alumino-silicate layers and the dye (i.e., the interaction between the electron lonepairs of clay surface oxygen and the dye -system) and (ii) the interaction between the dye molecules in the interlayer space of the clay structure (i.e., dimerization and the – interaction between two monomers in the dimer). Rhodamine dyes are considered good probe molecules to study the clay–dye complexes because they can be easily intercalated into the clay structure via a cation-exchange mechanism and their photophysics depends on the environmental factors. The molecular mechanics and dynamics simulations have been performed in Cerius2 to investigate the structure of montmorillonite intercalated with rhodamine B cation. Montmorillonite belongs to the 2:1 dioctahedral layer silicates with prevailing octahedral substitutions in the silicate layer. The octahedral and tetrahedral substitutions in the silicate layers lead to charge fluctuations in the surface oxygen sheet and the resulting charge distribution is important factor ruling the arrangement of guests in the intercalated structure. It is evident from the silicate layer structure that the charge fluctuations in the surface oxygen sheet due to the octahedral substitutions are lower than fluctuations due to the tetrahedral substitutions. Molecular modeling in Cerius2 shows that in case of montmorillonite layers with prevailing octahedral substitutions the charge on the surface oxygen sheet is almost evenly distributed with only small fluctuations within the range −0 03 el. Consequently there are no strong pronounced active sites for the anchoring of guests on the silicate layer. This leads to irregular anchoring of guests on the silicate layer and to disorder in layer stacking in structures of intercalated montmorillonites [349]. The starting model for building the montmorillonite layer is based on the structure data published by Tsipursky and Drits [84]. The unit cell parameters according to Méring and Oberlin [350] have been used to define the planar unit cell dimensions: a = 5 208 Å, b = 9 020 Å. The composition of the montmorillonite layer according to the chemical analysis was (Al1 53 Mg0 23 Fe3+ 0 25 (Si3 89 Al0 10 Ti0 01 O10 (OH)2 . To create a supercell of reasonable size for calculations the structure formula was slightly modified. Consequently the supercell 3a × 2b × 1c with the layer composition (Al18 Mg3 Fe3+ 3 (Si47 Al1 O120 (OH)24 was built with the total negative layer charge (−4). The c values in the initial model were set up according to the basal spacings obtained from X-ray powder diffraction [351, 352]. Modeling results led to two types of interlayer structure with the basal spacing 18.1 (phase 18 Å) and ∼23 Å (phase 23 Å) [351, 352]. The existence of these two phases was corroborated by experiment (X-ray powder diffraction). Both phases differ significantly in the interlayer structure; anyway they exhibit certain common features. First of all no

663 regular ordered positions of the rhodamine B anchoring to the silicate layers have been found, neither for phase 18 Å nor for 23 Å. The intramolecular rotation about the xanthene–amine bonds has been observed in both phases. The carboxyphenyl rings may rotate about the xanthene– phenyl bonds in both phases and in all arrangements of rhodamine B cations. The carboxyl groups can rotate about the phenyl–carboxyl (C–C) bonds. The slight distortion of the planar xanthene part of the rhodamine B cation has been observed in both phases 18 Å and 23 Å and confirmed by IR spectroscopy. Phase 18 Å (d001 = 181 Å) The rhodamine B cations form a bilayer arrangement. In both guest layers the long axis of the xanthene ring along the N–N line is nearly parallel with the silicate layers. The phase 18 Å exhibits two types of interlayer structures with the same basal spacing and nearly the same total sublimation energy: dimeric sandwich type and monomeric bilayer arrangement [351, 352]. Dimeric Sandwich-Type H-Dimers These were reported also in [208, 353] (see Fig. 20a, b). In H-dimers the rhodamine B cations interact with each other via the carboxyl groups. A detailed view of the H-dimer is seen in Figure 20b. One can see the carboxyl groups pointed to the oxygen atom in the xanthene ring of the neighboring rhodamine B cation. It is also evident from Figure 20b that the double xanthene = amine bonds of both cations in the dimer are oriented in the same direction (head-to-head arrangement). The total sublimation energy per one supercell of montmorillonite 3a × 2b × 1c is 876 kcal/mol.

(a)

(b)

Figure 20. (a) Montmorillonite intercalated with [rhodamine B]+ , phase 18 Å with bilayer dimeric arrangements of rhodamine B, Hdimers. (b) The detailed view of H-dimer intercalated in montmorillonite in the phase 18 Å.

Dye/Inorganic Nanocomposites

664 Monomeric Bilayer Arrangement This is shown in Figure 21a, b. Figure 21a shows the side view of the 3a × 2b × 1c supercell with monomeric bilayer arrangement of rhodamine B cations. The top view of the guest structure is given in Figure 21b. In this structure only weak interactions between the guests were detected. The basal spacing for this model is the same as for the H-dimeric structure 18.1 Å. The total sublimation energy per supercell 3a × 2b × 1c is slightly lower, ∼842 kcal/mol. The inclusion of the water molecules into the interlayer space of the phase 18 Å leads to a significant increase of the total sublimation energy. Thus it is very probable that the sorption of water is stabilizing the structure. This conclusion of modeling agrees with experimental observations. The modeling of monomeric arrangement in the interlayer space of montmorillonite led to: (i) the bilayer guest structure (phase 18 Å) previously described and (ii) the monolayer

arrangement of tilted monomers with basal spacing within 21–25 Å, which we denoted as “phase 23 Å” [351, 352]. Phase 23 Å This was observed by X-ray diffraction in samples prepared using intercalation solutions with very low concentrations of rhodamine B with prevailing monomeric arrangement. The modeling result of this structure is shown in Figure 22a. In this phase the rhodamine B cations are tilted with respect to the silicate layers. The tilting angles are in the range 40 –60 . This wide range of positions and orientations of rhodamine B cations in the interlayer results in disorder and strain in the interlayer space. There is a mixture of monomers and dimers in the interlamellar space. Two types of dimers observed are shown in Figure 22b: a head-to-tail sandwich type of dimer and a head-to-tail J-dimer. (For more details see [351, 352].) Insertion of water molecules into the interlayer structure of the phase 23 Å leads to the fluctuations of the basal spacing, within the range 21–25 Å.

(a)

5.4.3. Vermiculite Intercalated with Anilinium Vermiculite is a widely used host structure for the intercalation of various organic species. In contrast to intercalated smectites intercalated vermiculites exhibit ordered

(a)

(b)

(b)

Figure 21. (a) The structure of montmorillonite intercalated with [rhodamine B]+ , phase 18 Å with monomeric bilayer arrangement of guests. (b) The top view of the interlayer guest structure of [rhodamine B]+ in the phase 18 Å with monomeric bilayer arrangement of guests.

(a)

(b)

Figure 22. (a) Structure of montmorillonite intercalated with [rhodamine B]+ , phase 23 Å with monolayer arrangement of guests, which can occur in monomeric and dimeric arrangement (J-dimers, head-totail dimers, H-dimers, monomers). (b) Detailed view of the head-to-tail dimer (a) and J-dimer (b) in monolayer arrangement of the phase 23 Å.

Dye/Inorganic Nanocomposites

structures, which could be solved using single crystal diffraction [354–356]. Vermiculite is a 2:1 trioctahedral layer silicate. The layer charge in the vermiculite comes from the tetrahedral substitution (replacement of Si by Al). Compared with the trioctahedral saponite the layer charge in vermiculites is higher due to the higher concentration of tetrahedral replacements. The higher concentration of replacements and their tendency to separation leads to a higher degree of ordering in the silicate layer [357] and consequently to a more regular charge distribution in the surface oxygen sheet. This is a favorable condition for the ordering of guest cations in the intercalated structure. Due to the higher layer charge and consequently higher guest concentration the guest–guest interaction becomes more important for the guest arrangement in vermiculites than in smectites. The ordering of the guests in vermiculite is a result of interplay between the host–guest and the guest– guest interaction. The ordered structure of vermiculite intercalated with anilinium cations was investigated using modeling with an empirical force field in Cerius2 [358]. The results of modeling have been compared with the single crystal X-ray diffraction data published by Slade and Stone [356]. The initial model of anilinium–vermiculite was built using the structure data presented by Slade and Stone [356]: space group C2/m with the cell parameters a = 5 33 Å, b = 9 268 Å, c = 14 89 Å, and & = 97 . The host structure was Na-saturated Llano vermiculite from Llano County, Texas. The structural formula of the silicate layer was (Mg5 62 Al0 16 FeIII 0 13 Ti0 04 Mn0 01 (Si5 79 Al2 21 O20 (OH)4 . In order to create a unit cell of reasonable size for modeling, we had to simplify the structure formula of one unit cell, taking into account not only the layer charge but also the reported C and N contents in the intercalate. Two aniline cations have been placed into the single unit cell with the layer formula (Mg6 (Si6 Al2 O20 (OH)4 , carrying the charge −2 and corresponding to the fully exchanged vermiculite. The result of modeling is presented in Figure 23a, b. The anilinium cations are positioned over the ditrigonal cavities in the silicate layers to which the C–N bonds are perpendicular. The anilinium cations alternate in anchoring their terminal NH+ 3 groups to the lower and upper silicate layers. The three amine–hydrogen atoms of aniline are involved in hydrogen bonds to the surface oxygen in adjacent silicate layers. The arrangement of anilinium cations in the interlayer space and the mutual positions of two successive silicate layers are illustrated in Figure 23b. This arrangement of anilinium cations in the calculated model corresponds very well to the electron density map obtained from the single crystal diffraction data, by Slade and Stone [356]. A comparison of the calculated cell parameters with the experimental structure data of Slade and Stone [356] is given in Table 9. The ditrigonal cavities in adjacent silicate layers are positioned opposite each other and this order in layer stacking is controlled by the well-defined positions of the anilinium cations. This regular layer stacking has also been confirmed by X-ray single crystal diffraction measurements in the Slade–Stone model [356]. The average N–O distance (averaged over six oxygens surrounding the adjacent ditrigonal cavity) in our calculated model dN−O is 3.01 Å; in the experimental Slade–Stone model dN−O = 3 03 Å.

665 (a)

(b)

Figure 23. (a) Vermiculite intercalated with anilinium. Hydrogen bonds between the anilinium hydrogen and surface oxygen on the silicate layer are marked by a broken line. (b) The regular arrangement of anilinium in the interlayer space of vermiculite. The top view illustrates the regular stacking of the host layers. Two adjacent tetrahedral sheets are visualized as polyhedra (lower silicate layer) and cylinders (upper silicate layer).

In a subsequent paper Slade et al. [359] discussed the following problems with the structure determination. (a) The packing density of the anilinium is not as high as expected from the structural model presented earlier and the ion exchange capacity of the Llano vermiculite. (b) Therefore, there are additional water and inorganic cations in the structure (anilinium content too low with respect to the cation exchange capacity). However, it is difficult to Table 9. Comparison of cell parameters for anilinium–vermiculite.

Experimental Modeling

c Å

  

&  

d Å

14.89 14.84

90.00 90.53

97.00 97.17

14.78 14.72

Note: Experimental values have been taken from X-ray single crystal diffraction data of Slade and Stone [157]. The calculated values have been obtained by modeling using Cerius2 [159].

666 determine the absolute positions of these inorganic cations, the water molecules, and the organic cations. Based on three-dimensional data the author introduces the simultaneous presence of anilinium rich and anilinium pure regions to overcome these problems. The first one shows the arrangement of aniline as discussed in the previous paper [356] and in this chapter. This packing motive is also found for benzidine ions in Young River (Western Australia) vermiculite [360]. In the anilinium poor domains triangular groups of water molecules about the inorganic cations and the anilinium ions form together a close-packed array. The anilinium loaded Llano vermiculite is dark and lilacblue colored indicating that anilinium has been oxidized by the low amount of structural iron [360]. Since the anilinium ions in the interlayer space are highly organized, the question arises as to where the oxidation takes place and what the oxidation products are. One explanation could be that the color arises from anilinium radicals and the arrangement of the anilinium ions prevents the oligomerization. Another possibility is that the oxidation is only a side reaction of the formation of the intercalation complex and it takes place on the external crystal surfaces and the products remained there because they are too big to diffuse in the interlayer space.

5.4.4. [Ru(bipy)3 ]2+ /Saponite As pointed out the [Ru(bpy)3 ]2+ complex is the most studied photosensitizer in biomimetic artificial photoelectrochemistry and a widely used electron transfer mediator in organic electrochemistry. Apart from that Yamagishi has shown that some chiral transition metal (mainly Fe2+ , Ni2+ , Ru2+  complexes with chelate ligands like phenanthroline, bipyrdine, or 1-(2-pyridylazo)-2-naphthol intercalated in low charged smectites exhibit remarkable chiral recognition phenomena despite the fact that the silicates are achiral [165]. Thus, a racemic mixture of the [Fe(phen)3 ]2+ complex is taken up in montmorillonite in an amount twice as high as the cation exchange capacity of the clay and twice as high as the pure enantiomers. The layer expansion of the product is in the order of 20 Å indicating the uptake of metalcomplex double layers, whereas only a monolayer of the pure enantiomer complexes is adsorbed. In case of the [Ru(bpy)3 ]2+ the opposite effect is observed: bilayer adsorption of the pure enantiomer complex and monolayer adsorption for the racemic mixture [361]. The preferential uptake of racemic pairs is confirmed further by the observation that the addition of clay to a partially resolved mixture of both enantiomers changes the optical purity of the solution after filtration [165]. In addition the following other effects has been discovered: induction of optical activity in a less stable chiral chelate complex by preadsorption of an enantiomeric chelate complex, chromatographic separation of enantiomeric pairs with clays loaded with an enantiomeric chelate complex, and asymmetric synthesis. However, the reason for this preferred formation of racemic pairs is still a matter of debate [362], since it was not possible to determine the structure of a clay loaded with racemic or enantiomeric chelate complexes. It is clear that the complexes are arranged in the interlayer space with their pseudo C3 symmetry axis perpendicular to the clay layers. The protons of the aromatic rings are keying in three neighboring

Dye/Inorganic Nanocomposites

ditrigonal cavities on the Si–O tetrahedral network [165, 362]. Modeling of the possible arrangements of [Ru(bpy)3 ]2+ in a (saponite) gave different packing motifs for the pure enantiomer and the racemic complex. Whereas the enantiomer shows the generally assumed hexagonal pattern, the arrangement of the racemic mixture is of tetragonal symmetry. The distance between complexes of the same chirality is slightly shorter along the b-axis and the neighboring molecules penetrate each other. This interpenetration of the optical antipodes is feasible only due to the clay mediated alignment of the complexes (parallel C3 axis) [362] and is indicative for a stronger guest–guest interaction. This result confirms the peculiar interactions within a racemic pair, but it is in disagreement with the assumed packing for the Fe–phenanthroline complexes and with the assumption that racemic pair formation has to be accompanied by the formation of a double layer arrangement [165]. This structure simulation is performed for a loading level high enough to yield a close packed monolayer. For designing microporous materials it is necessary to know how reducing the layer charge will affect the interlamellar arrangement of the Ru complexes. In a subsequent investigation the same authors clarified this point and highlighted the complexity of the interplay of the host–guest and guest–guest interactions [349]. In this study the overall charge of the host is reduced by more than a factor of two with respect to the previous study (0.22 per formula unit vs 0.5). For the [Ru(bpy)3 ]2+ intercalated into a homogeneously charged saponite (negative charges due to substitution in the tetrahedral layers) the energy minimum for both enantiomeric and racemic interlayers is observed with the hexagonal arrangement of the complexes (E1 and R1 in Fig. 24). The picture changes when switching to the very similar pillar [Ru(phen)3 ]2+ . For enantiomeric interlayers E1 is still energetically favored over any other cluster investigated. However, with the racemic complexes -stacks can be realized between aromatic ligands of neighboring cations. Due to the parallel alignment of the C3 -axes, as required by the host–guest interaction, the cation distances and orientations induced by the clay surface corrugation the guest–guest interactions are stronger for larger -systems (phen as compared to bpy) and even large clusters like one-dimensional chains became lower in energy than the hexagonal pattern (R4 in Fig. 24). Moving the permanent silicate charge from the tetrahedral to the octahedral layer and hence further away from the interlayer cations increases the weight of the guest–guest interaction, which makes clustering more feasible. The difference of E1 and E4 declines when going from saponite to hectorite. For the racemic mixture pairs of [Ru(bpy)3 ]2+ (R2) have energies so close to that of the hexagonal pattern that ordering cannot be ruled out. In case of the enantiomeric interlayers the influence of a nonrandom charge distribution in the host structure has been investigated. Two charge models have been compared. In one of them the substitution follows the pillars in model E4; in the second one it reflects the maximum Löwensteinian clustering of isomorphous substitution. Starting simulations with a prevalent mismatch between the isomorphous substitution pattern in the host and the pillar arrangement in the interlayer are far from any local energy minimum. On the other hand, configurations where pillar arrangement and substitution pattern

Dye/Inorganic Nanocomposites

R1

667

E1

neously charged host materials in order to be able to construct highly organized intercalation compounds. Therefore when working with natural smectites, the characterization of exchange properties solely by bulk cation exchange capacities will make a conclusive interpretation and reproduction difficult. This point can be illustrated by the conflicting results of luminescence behavior of racemic and enantiomeric [Ru(bpy)3 ]2+ and [Ru(phen)3 ]2+ intercalated in smectites [363–366].

5.5. Examples of Structure Analysis in Framework Structures

R2

E2

R3

E3

R4

E4

R5

R6

Figure 24. Possible arrangements of [Ru(bpy)3 ]2+ /[Ru(phen)3 ]2+ confined between homogeneously charged saponite. The configurations correspond to local energy minima for the enantiomeric [E] and racemic (R) complexes [349].

correspond have local minima close to the starting structures. Despite the increased clustering between both models the pillar arrangement does not change significantly because short-range guest–guest and/or host–guest interactions do not allow denser packing. These results put the focus on the following points. (a) The electrostatic energy is the leading term in host–guest interactions. (b) Constructing supercells with a homogeneous substitution pattern creates highly organized structure in the interlayer. (c) Unless there is good reason to assume such an ordering, this approach can introduce severe artifacts; it will be essential to know the charge distribution and to synthesize monophasic, homoge-

As for the layered host–dye complexes the knowledge of dye arrangement in framework structures is rather limited. The structure determination has been carried out by X-ray powder diffraction, Rietveld refinement, and structure simulation with commercial programs (Cerius2 and ChemX); only in case of p-nitroaniline (PNA)/H-ZSM-5 was a single crystal structure analysis possible. The position of the guest molecules depends on the preparation method, the size and shape of the molecules inserted, and the pore structure in the frameworks. If the available space is too big, the inserted dyes are in close contact to the channel walls (methyl viologen in zeolite L). When inserting the species during synthesis of the framework, they are squeezed in the open windows between neighboring supercages (methylene blue and thioindigo in zeolites X/Y). Even the deformation of the framework cannot be excluded (p-nitroaniline in H-ZSM-5). In none of these examples is there a remarkable guest–guest interaction, which plays an essential role in layered intercalation compounds.

5.5.1. H-ZSM-5 Loaded with p-Nitroaniline If a crystal of H-ZSM-5 is loaded with PNA coexisting domains of different symmetries are formed depending on the reaction time [367]. At very long loading times crystals containing only one type of domain can be prepared. The PNA molecules are located in the intersections of the straight and the sinusoidal channels [367]. Figure 25c shows the orientation of PNA along the straight channels. The long axis of the molecules is arranged parallel toward the channel axis (crystallographic b-axis), but tilted by about 9 with respect to it. The angle between the normal on the aromatic ring plane and the a-axis is around 45 . Due to the adsorption of the PNA the straight channel pores are strongly elliptically deformed while the sinusoidal channels are nearly not affected by the adsorption [compare structures of the pristine ZSM-5 (Fig. 10) and the loaded system (Fig. 25a)]. This structure refinement is based on the centrosymmetric space group Pnma and not the acentric Pn21 a assumed in an earlier paper [369]. This increase in symmetry does not affect the position and orientation of the molecules seriously. Neighboring molecules are only loosely connected as can be seen in the N · · · O distances of about 4 Å, which is too long for hydrogen bonds [367]. The nitro group is shifted away from the straight channel-10 ring and preferentially related to the neighboring molecules by a twofold screw axis along b (non-centro-symmetric structure), to avoid the short O · · · O contact. If this chain is interrupted by empty

Dye/Inorganic Nanocomposites

668 (a) metyl viologen in zeolite L

(b) thioindigo in a faujasite type zeolite

thus it is transferred to trimethylthionine. The incorporation of MB/trimethylthionine is accompanied with a complete loss of Na at SI sites (cf. Fig. 10, top). Whereas MB/trimethylthionine is located in the center of the effective window between neighboring supercages, the position of thioindigo depends on the condition of solvent removal. By a mild removal of solvents (in air at room temperature) the molecule is located in the center of the window. The occupation factors, exhibiting values PMMA > sol–gel bulk [460].

6.4. Second Harmonic Generation Second harmonic generation is a nonlinear optical process that converts an input optical wave into an out wave of twice the input frequency. To gain insight into the origin of nonlinear optical effects it is helpful to look at the nature of

683 polarization induced in a molecule by a local electric field [461]. The polarization P can be expended in powers of the electric field E, P = E + &E 2 + %E 3 + · · ·

(5)

The first term  in the expansion is the linear polarization and is the origin of the refractive index if the field E is associated with an electromagnetic wave in the optical frequency range. The coefficients are actually complex numbers. The real part of  corresponds to the index of refraction, and the imaginary part corresponds to absorption of a photon by the molecule. When an electromagnetic field interacts with a medium consisting of many molecules, the field polarizes the molecules. These, in turn, act as oscillating dipoles broadcasting electromagnetic radiation, which can then be detected at some point in space outside of the medium. In a nonlinear medium the induced polarization is a nonlinear function of the applied field and is related to the nonlinear term & (its real part is the molecular hyperpolarizability), which makes the most significant contribution to the induced frequency components. It is nonzero only in noncentro-symmetric media. (% is the first nonzero nonlinear term in centrosymmetric media.) If the response P is asymmetric, an appropriate summation of the even harmonics describes the function P (SHG). Nonlinear optical properties are measured on macroscopic samples that consist of many individual molecules. Although it is possible to infer the values of the molecular hyperpolarizabilities from measurements on macroscopic samples, great care must be taken to determine the values of the internal electric fields and the propagation characterisitcs of the generated fields. In the design of materials with large second-order nonlinear optical coefficients and other useful properties the molecular hyperpolarizabilities must be optimized and then oriented in a medium so that the propagating field can be transmitted from the medium. Recent studies have revealed large molecular hyperpolarizabilities of certain organic materials (see Section 3.3) leading to anomalously large optical nonlinearities relative to those of conventional inorganic substances (like LiNbO3  [161, 162, 461]. Major research efforts are directed toward (a) identifying new molecules possessing large nonlinear polarizability and (b) controlling the molecular orientation on a microscopic level. The following part of this section will be devoted to the contribution of intercalation in inorganic framework to the latter point. Cooper and Dutta have reported the first evidence for SHG in a layered intercalation compound [462]. They have inserted 4-nitrohippuric acid (NO2 C6 H4 CONHCH2 · COOH = NHA) into the layered lithium aluminate [LiAl2 (OH)6 ]+ (see Fig. 8). From the observed basal spacing of the product (d = 18 Å) the guest species are thought to be stacked perpendicular to the host lattice sheets. From optical spectroscopic data it has been concluded that the NHA molecules are held in the interlayer space as carboxylic acids. The material exhibits frequency-doubling characteristics generating a 532 nm radiation from incident 1064 nm. This is particularly interesting, since crystals of pure NHA do not exhibit any nonlinear optical properties due to the centrosymmetric packing of the crystals. Thus,

684 the dipoles of the NHA molecules must be aligned very strongly and the arrangement must be very particular and non-centro-symmetric, while the host lattice itself is also centrosymmetric. Lacroix et al. have reported SHG in intercalation compounds of DAMS+ (see Section 3.3.6, Table 7 and Scheme 6) in MPS3 , where M is either Mn2+ or Cd2+ [249, 250]. In their first report on this system the authors prepared this compound by using the potassium intercalation compound K2x (H2 O)y Cd1−x PS3 as an intermediate. This product gave only a small SHG response on the same order as that of urea, whereas the DAMS+ chromophore can be 103 times more efficient than urea. This disappointing poor SHG signal was thought to originate from grain surface effects and low crystallinity of the intercalation compound. The authors could optimize the synthesis by direct reaction between MPS3 powder and an ethanolic solution of DAMS+ at 130  C in the presence of pyridinium chloride. Pyridinium chloride played a role in generating in-situ an intermediate pyridinium intercalation compound that underwent rapid exchange with DAMS+ . The products obtained had the composition (DAMS)0 28 Cd0 86 PS3 and (DAMS)0 28 Mn0 86 PS3 , respectively, and exhibited a significantly increased SHG activity (750 and 300 times, respectively, that of urea) [249]. In a more thorough investigation of the optical properties of MPS3 intercalation compounds with various stilbene derivatives it could be shown that only those compounds show a significant SHG effect, which form J-aggregates (see Section 6.1 and Fig. 26; the adjacent DAMS+ molecules are oriented with their long axis parallel to the host layer planes and the benzene rings perpendicular to the MPS3 sheets) [251, 429]. This is the case nearly exclusively for the two aforementioned compounds. Since the original host lattice is centrosymmetric there must be a very dense packing of the DAMS+ cations within a non-centro-symmetric superstructure. However, X-ray studies did not give evidence for such a superstructure or other structural changes from one intercalate to the other, other than small differences in the host interlayer space [429]. Thus, there seems to be no longrange order, but for an effective SHG to occur aggregates or ordered domains of a sufficient size of coherence (around 1 m) should be present. For the detailed characterization of the optical properties as well as their practical applications as photonic devices, the compounds should be available as single crystals or oriented thin films. Such an assembly has been realized with an organic Zr–diphosphonate prepared by means of the layer-by-layer method (see Fig. 16) [312]. Instead of the hemicyanine chromophore in the aforementioned compound an azobenzene moiety is used (cf. Table 7 [259, 260], Scheme 5). The chromophores with asymmetric head and tail groups (to produce a high dipole moment) are oriented perpendicular to the inorganic layers. The nonlinearities of this system are on the same order of magnitude as those of LiNbO3 , one of the widely used inorganic nonlinear materials. The SHG shows that the multilayers have polar order that does not decrease with increasing numbers of monolayers in the film. The inorganic interlayers impart excellent orientational stability to the dye molecules, with the onset of orientational randomization above 150  C. Relatively large values for some of the hyperpolarizability tensors indicate that it is inappropriate

Dye/Inorganic Nanocomposites

to view this dye as an isolated linear molecule with a single hyperpolarizability along its long axis. The polar order of the films is greater using a rigid chromophore than with a more flexible one [260]. However, SHG activity can be lost through H-aggregation, if the component molecules are both highly ordered and very polar [260]. (The other donors used in [260] are mentioned in Table 7.) The SHG in the film using the phosphonate shown in Scheme 5 is not well understood. The film does not exhibit linear dichroism and X-ray diffraction studies indicate the absence of a well-structured surface [259, 461]. In addition, ellipsometry studies of the interlayer system on silicon indicate that each deposited layer is approximately 16 Å thick, which is 4 Å thinner than calculated based on typical bond lengths [312]. The difference in layer thickness was attributed to nonideal packing and an average molecular tilt angle of ∼35 . Some of these inconsistencies may be due to loss of Zr4+ from the surface and some “self-healing” during subsequent exposure to layer formation solutions [463]. (Note also the tendency for hydrolysis of the Zr–phosphate layers [450]!) That the film formation is not as ideal as generally assumed is further confirmed by angle-resolved photoacoustic spectroscopy [259]. The curvature of the photoacoustic amplitude as function of the electric field angle increases with increasing number of deposited layers, indicating a change in molecular orientation as a function of coverage. The authors assume that incomplete interlayers are formed during each exposure and that subsequent exposures result in improved coverage for the bottom layer as well as formation of an incomplete upper layer. SHG has been observed also for PNA and related aromatic polar molecules (Table 8) inserted in aluminophosphate molecular sieves (AlPO4 -5, AlPO4 11, VPI-5). Models supported by Fourier transform infrared evidence shows that the aggregates could be chains of molecules with a large net dipole that form within the molecular sieve channels. The polar host then aligns the aggregates to produce a large bulk dipole moment and the observed SHG signal [8, 268] if the SHG effect of PNA/AlPO4 becomes significantly larger when measured on a single crystal sample or in crystals aligned in the electric field [322]. Measurements of the pyroelectric effect show that the PNA dipoles are oppositely directed into halves of a single crystal, forming two macroscopic regions of opposite polarization. It is concluded that the preferred direction of the dipole chains in the channel is a result of the adsorption process of the molecules in the channels [269]. However, because of the macroscopic extent of each of the two regions (comparable with half of the crystal length) being much larger than the wavelength of the incident laser light, the second harmonic order light from both regions does not interfere [322]. The PNA molecules adsorbed in ZSM-5 also exhibit the SHG effect. This is not a trivial finding, since in contrast to the AlPO4 -5 case where reorientations are excluded for steric reasons, the bimodal and bidirectional MFI framework could a priori allow reorientation of the PNA molecules in the channel intersections [368]. The angular dependence of the SHG signal together with the position of the ZSM-5 crystal in the laser beam show that the SHG signal detected originates exclusively from PNA molecules that are aligned parallel with their long molecular axes to the straight channels [368].

Dye/Inorganic Nanocomposites

X-ray single crystal structure analysis is in good agreement with these findings (see also Section 5.5.1) [367, 369]. The PNA molecules are located in the intersections of the two interconnected channel systems. However, there is disagreement about the symmetry of the composite. In the previous paper the crystal structure has been solved in a non-centrosymmetric space group, with symmetry lower than that of the empty host lattice [369]. This result could not be confirmed later [367]. The loss of acentricity is explained by empty intersections, which destroy the long-range coherence of the polar chains [367]. Maybe it is enough for SHG that there are ordered non-centro-symmetric domains with a size larger than the wavelength of the incident light, as seems to be the case in other SHG materials discussed in this section. PNA has been intercalated also in tetramethylammoniumsaponite. The free space between the TMA ions is just as large as the PNA molecules [464]. These additional guests are oriented with their long axes parallel to the silicate layers. During thermal intercalation the dipoles are oriented randomly in the ab plane and do not show an apparent SHG effect. However, when the intercalation has been carried out in an applied electric field, the sample exhibites an SHG effect with an intensity 105 higher than that for a compound without electric field. This indicates that the applied electric field caused the alignment of PNA dipoles with noncentro-symmetry in the interlayer space. In this experiment the SHG effect is used as analytical tool to monitor the orientation of the highly polar molecules. For dimethylaminobenzonitrile (DMABN) in the straight pores of AlPO4 -5 two components of the SHG effect have been observed [270]. One component of the frequencydoubled radiation is polarized to the long axis of the crystal (z-axis, *2 = 0 ) and exhibits the maximum of intensity when the excitation beam has the same polarization. The second component is polarized perpendicularly to the z-axis (*2 = 90 ) and shows the maximum efficiency for mixed excitation (*2 = 45 ). The reason for the different tensor elements is the two close-lying lowest excited states in DMABN having transition dipole matrix elements perpendicular to each other. Whereas PNA loaded molecular sieves are able to mix waves of the same polarization only, for the DMABN-based materials a mixing of waves with different polarizations is observed. This is a promising property for phase matching as a prerequisite for technical applications. The investigations described were mainly motivated by the search for new materials for nonlinear optics. However, all the results give insight in the orientation of molecules in solid matrices. Two papers, which remain to be discussed, are devoted exclusively to showing application of the SHG as an analytical tool in surface chemistry. Yan and Eisenthal reported a SHG signal in aqueous suspension of clay particles loaded with 4-(2-pyridylazo)resorcinol (PR) [261]. Since montmorillonite has a centrosymmetric structure, SHG cannot be generated from the bulk of the sample. Because of the coherence properties of SHG, the SHG efficiency greatly depends on the size of the particles. Since the diameter of the clay particles is on the length scale of the light wavelength, while the thickness of the particles is much smaller than the wavelength, only the edges separated by roughly the diameter of the particles contribute to the SHG effect [261]. The origin of the SHG effect in the clay suspension must

685 be the polarization of water molecules by surface charges located at the edges of the clay particles. On addition of PR to the clay aqueous dispersion a strong SHG signal originating from PR molecules adsorbed onto the edge surfaces was observed. The nonlinear optical properties of a glass/clay/PDDA/ NAMO films were found to depend on the adsorbed amount of NAMO (a negatively charged phenylazobenzenesulfonic acid derivative) and its non-centro-symmetric organization [311]. These factors are in turn governed by the substrate type, the clay type, and the PDDA (poly-diallyldimethylammonium) concentration. Optimized SHG for the glass/laponite/PDDA/NAMO films was found, when the clay articles are deposited on a (3aminopropyl)trimethoxysilane modified glass surface and PDDA chains are adsorbed from a 0.1 M solution. At high PDDA loading the length of unbound polymer strands provides enough structural flexibility for the dipole moments of the dye. The overall non-centro-symmetric organization decreases together with the SHG intensity.

7. CONCLUSIONS AND FUTURE PERSPECTIVES • Despite the huge number of dye/inorganic composites studied only a few of them found greater attention and are promising for applications: the [Ru(bpy)3 ]2+ complex as a photosensitizer (in organic supramolecular assemblies and in nature porphyrins do that job), methyl viologen (in natural systems quinines) as the most widely used electron acceptor, rhodamines in lasing, azobenzenes, spiropyranes, and viologens (perhaps) for photochromism, and azobenzenes and stilbenes for SHG. This is normal in many other areas of applied chemistry (for example, 1 out 10,000 compounds is pharmaceutically active and applicable; 5 out of 5000 compounds are applied superconductors). • Many of the dyes discussed in this chapter show the tendency for aggregate formation. For many applications this is an unfortunate property because it reduces the quantum efficiency of luminescence, reduces the lifetime of the excited state, and influences energy and electron transfer. Encapsulation in organic or inorganic matrices is the method to circumvent these unfortunate side reactions. • Framework structures with channel diameters of the dye size provide the constrained space for preventing aggregate formation, restricted motion, and an alignment of the dye molecules. If alignment is the essential property as is the case in application in nonlinear optics dye/zeolite composites are a good choice. If the contact between neighboring dye molecules has to be reduced as in the formation of light antennas zeolite encapsulated dyes have some advantages. If free space is of advantage as in photoisomerization processes (photochromism) encapsulation into gels is an interesting alternative. • Despite the perspectives for application photophysical investigations are valuable tools for studying guest– guest and guest–host interactions, even in systems with missing long-range order in the dye arrangement.

Dye/Inorganic Nanocomposites

686 • The knowledge of the arrangement of dye species is rather limited despite the huge number of investigations of dye/inorganic composites. Lack of long-range order, disorder, and mobility lead to enormous difficulties in structure determination by diffraction methods. The situation seems better in the case of the highly crystalline molecular sieves and in vermiculites. Note that most structural data of intercalated organic compounds are obtained for vermiculite intercalation compounds. The situation may improve further when applying synthetic clays for structural work [465, 466]. Since the middle of the 1990s highly sophisticated new tools for structure simulation have gained more and more importance in structure determination and will support strongly structure determination. • Despite the demonstration that artificial multilayer assemblies can be obtained, knowledge about reproducibility of assembly preparation, comparability of the physical properties in various preparation runs, and stability of highly sophisticated assemblies is rather low. The structural view of the artificial composites seems to be very idealized. Most of the systems discussed will be metastable states far from equilibrium. Many questions arise concerning the kinetic stability, the degree of disorder, the types of defects, and their annealing behavior at room temperature. • Great progress has been made in orientation of zeolite crystals. • Whereas layered compounds, especially clays, were the first host lattices for the preparation of dye/inorganic composites, and a wealth of knowledge about intercalate arrangement is now available, the products are far removed from applications of dye/zeolite and dye/gel composites. This is due to the very strong guest–guest and guest–host interactions and the missing organization of the intercalation compounds. It is worth mentioning that most of the dye/layered host composites are prepared via flocculation of colloid dispersion; thus badly organized tactoids are formed. In addition, the packing density of the dyes in the interlayer space and the tendencies for aggregation detracted from increases in quantum efficiency and optimization of other relevant optical properties. • To keep the layered materials in the business it is crucial to control the coadsorption/cointercalation of photosensitive electron donors, electron acceptors, and other functional molecules. Application of the concepts of organic supramolecular chemistry [435, 467– 471] may help one to find better assemblies. However, one should have in mind that for mimicking, say biological photolysis/photosynthesis, it may be necessary to arrange different sorts of antenna molecules and electron transfer shuttles in a highly organized but flexible arrangement, probably all of the species in very different concentrations (cf. the most recent success in constructing light antennas on the basis of dye/zeolite composites [410, 411, 432, 433]). Structure directing chromophores as used in the hydrothermal in-situ syntheses of dye inorganic composites may be of advantage for this goal. Suitably modified molecules could be arranged in the constrained space of solids and the

functional units constructed in this frame (ship in a bottle synthesis). Another interesting aspect is the formation of amphiphilic organic–inorganic heterostructures as prepared for the first time by Jido et al. [472–474]. Breu et al. obtained highly ordered intercalation compounds and heterostructures with synthetic fluorohectorites [465, 466]. Cointercalation of surfactants may also allow a better organization of intercalated species, if demixing can be avoided [14, 475]. The organization of the platelets [248] and even of the intercalated species [464] could be improved also by the application of an electrical field.

GLOSSARY Classical molecular dynamics Introduces the kinetic energy into the system, where the temperature and velocities are related via Maxwell–Boltzmann equation. Potential energy is described using empirical force field. Empirical force field The set of empirical force constants, parametrizing the bonding (bond stretching, bending, torsion, deformation etc.) and nonbonding interactions (van der Waals, H- bonds, etc.), which are described using simple analytical expressions. Host-guest complementarity The set of geometrical and chemical factors characterizing the ability of both molecules to create an ordered inclusion intercalated structure. Intercalation The reversible uptake of atoms, ions, molecular cations or molecules at low temperature while the structure of the host lattices is conserved. The term “intercalation” does not tell anything about the nature of the chemical reaction underlying this process. It stresses mainly the topological relationships between the host and the final product of the reaction. Laugmuir–Blodgett technique Produces organised arrangements of amphiphilic molecules on a solid substrate. Layer-by-layer-deposition A preparative method applied for sequential adsorption of alternatively charged polyelectrolytes. It can be adapted to the sequential layering of inorganic polyanions (obtained by exfoliation of layered intercalation compounds: clays, ternary oxides, Zrphosphates) with a variety of polymeric cations. Molecular mechanics In molecular mechanics the potential energy of the system is described using empirical force field. Molecular simulations The optimization of structure and bonding geometry based on energy minimization. Photochromism A reversible transition between two states with different light absorption spectra, where the transition should be induced by light at least in one direction. Photodecomposition of water The light induced splitting of water for the production of hydrogen and/or oxygen. It is the energy supplying process for the assimilation of CO2 in photosynthesis. Second harmonic generation A nonlinear optical process that converts an input optical wave into an out wave of twice the input frequency. Sensitizer A light sensitive dye, which can transfer in the excited the electron to an electron acceptor.

Dye/Inorganic Nanocomposites

REFERENCES 1. W. Noll, Fortschr. Miner. 57, 203 (1979). 2. D. Wildung, Ägyptische Staatssammlung München. 3. F. Seel, G. Schäfer, H.-J. Güttler, and G. Simon, Chem. Zeit. 8, 65 (1974). 4. Very interesting with respect to Mayan blue is the following homepage: http://www.azulmaya.com. Part of the information given in this homepage concerning the X-ray crystal structure refinement has now been published; C. Chiari, R. Giustetto, and G. Ricciardi, Eur. J. Miner. 15, 21 (2003). 5. C. Reyes-Valerio, “De Bonampak al Templo Mayor, El Azul Maya en Mesoamérica.” Siglo XXI editores, Mexico, 1993. 6. M. José-Yacamán, L. Rendón, J. Arenas, and M. C. Serra-Puche, Science 273, 223 (1996). 7. J. K. Thomas, Acc. Chem. Res. 21, 275 (1988). 8. G. D. Stucky, Progr. Inorg. Chem. 40, 99 (1992). 9. J. K. Thomas, Chem. Rev. 93, 301 (1993). 10. K. B. Yoon, Chem. Rev. 93, 321 (1993). 11. S. Suib, Chem. Rev. 93, 803 (1993). 12. V. Ramamurthy, R. G. Weiss, and G. S. Hammond, Adv. Photochem. 18, 67 (1993). 13. J. Caro, F. Marlow, and M. Wübbenhorst, Adv. Mater. 6, 413 (1994). 14. M. Ogawa and K. Kuroda, Chem. Rev. 95, 399 (1995). 15. R. A. Schoonheydt, in “Comprehensive Supramolecular Chemistry” (G. Alberti and T. Bein, Eds.), Vol. 7. Pergamon, Oxford, 1996. 16. P. Behrens and G. D. Stucky, “Comprehensive Supramolecular Chemistry” (G. Alberti and T. Bein, Eds.), Vol. 7. Pergamon, Oxford, 1996. 17. V. Ramamurthy and M. A. Garcia-Garibay, “Comprehensive Supramolecular Chemistry” (G. Alberti and T. Bein, Eds.), Vol. 7. Pergamon, Oxford, 1996. 18. P. K. Dutta and M. Ledney, Progr. Inorg. Chem. 44, 209 (1997). 19. G. Schulz-Ekloff, D. Wöhrle, B. van Duffel, and R. A. Schoonheydt, Microporous Mesoporous Mater. 51, 91 (2002). 20. R. E. Grim, “Clay Mineralogy,” 2nd ed. McGraw–Hill, New York, 1968. 21. B. K. G. Theng, “The Chemistry of Clay-Organic Reactions.” Hilger, London, 1974. 22. E. A. Hauser and M. B. Leggett, J. Amer. Chem. Soc. 62, 1811 (1940). 23. D. Krüger and F. Oberlies, Ber. 74, 1711 (1941). 24. H. Weil-Maherbe and J. Weiss, J. Chem. Soc. 2164 (1948). 25. D. H. Solomon, B. C. Loft, and J. D. Swift, Clay Miner. 7, 389 (1968). 26. D. T. B. Tennakoon, J. M. Thomas, and M. J. Tricker, J. Chem. Soc. Dalton Trans. 2211 (1974). 27. M. M. Mortland and T. J. Pinnavaia, Nature 229, 75 (1971). 28. Y. Soma, M. Soma, and I. Harada, J. Phys. Chem. 89, 738 (1985). 29. T. Tanabe, “Solid Acids and Bases.” Kodansha/Academic Press, Tokyo/New York, 1970. 30. R. C. Mielenz and M. E. King, Proc. Am. Soc. Test. Mater. 55, 1213 (1951). 31. D. H. Solomon and D. G. Hawthorne, “Chemistry of Pigments and Fillers.” Wiley, New York, 1983. 32. S. Yariv, in “Organo-Clay Complexes and Interactions” (S. Yarif and H. Cross, Eds.). Dekker, New York, 2002. 33. P. Enzel and T. Bein, J. Phys. Chem. 93, 6270 (1989). 34. A. Moreale, P. Coos, and C. Badot, Clay Miner. 20, 29 (1985). 35. V. Mehrotra and E. P. Giannelis, Solid State Commun. 77, 155 (1991). 36. M. Ilic, E. Koglin, A. Pohlmeier, H. D. Narres, and M. J. Schwuger, Langmuir 16, 8946 (2000). 37. T. H. Chao and H. A. Erf, J. Catal. 100, 492 (1986). 38. T. Bein and P. Enzel, Angew. Chem. Int. Ed. Engl. 28, 1692 (1989).

687 39. Y. Soma, M. Soma, Y. Furukawa, and I. Harada, Clays Clay Miner. 35, 53 (1987). 40. P. Enzel and T. Bein, J. Chem. Soc. Chem. Commun. 1326 (1989). 41. J. V. Caspar, V. Ramamurthy, and D. R. Corbin, J. Am. Chem. Soc. 113, 600 (1991). 42. V. Balzani, L. Moggi, M. F. Manfrin, F. Bolletta, and M. Gleria, Science 189, 852 (1975). 43. C. R. Bock, T. J. Meyer, and D. G. Whitten, J. Am. Chem. Soc. 97, 2909 (1975). 44. C. Creutz and N. Sutin, Inorg. Chem. 15, 496 (1976). 45. C. T. Lin and N. Sutin, J. Phys. Chem. 80, 97 (1976). 46. J. M. Lehn and H. P. Sauvage, Nouv. J. Chim. 1, 449 (1977). 47. J. M. Lehn, H. P. Sauvage, and R. Ziessel, Nouv. J. Chim. 3, 423 (1979). 48. M. Calvin, Acc. Chem. Res. 11, 369 (1978). 49. H. Nijs, J. J. Fripiat, and H. van Damme, J. Phys. Chem. 87, 1279 (1983). 50. H. Van Damme, F. Bergaya, and D. Challal, in “Homogeneos and Heterogeneous photocatalysis” (E. Pelizetti and N. Serpone, Eds.), Reidel, Dordrecht, 1986, p. 479. 51. J. Kiwi, K. Kalyanasundaram, and M. Graetzel, Structure Bonding 49, 37 (1982). 52. I. Willner, J. M. Jand, C. Laane, J. W. Otwos, and M. Calvin, J. Phys. Chem. 85, 3277 (1981). 53. P. Sitte, H. Ziegler, F. Ehrendorfer, and A. Bresinsk, “Lehrbuch der Botanik für Hochschulen.” Gustav Fischer Verlag, Stuttgart, 1991. 54. C. H. Chang, D. Tiede, J. Tang, U. Smith, J. Norris, and M. Schiffer, FEBS Lett. 205, 82 (1986). 55. P. K. Ghosh and A. J. Bard, J. Am. Chem. Soc. 105, 5691 (1983). 56. A. Fiche, Clays Clay Miner. 38, 391 (1990). 57. A. Pockels, cited in W. Ostwald, Kolloid-Z. 58, 1 (1932); Lord Rayleigh, Philos. Mag. 48, 337 (1899). 58. I. Langmuir, J. Am. Chem. Soc. 39, 1848 (1917). 59. K. B. Blodgett, J. Am. Chem. Soc. 57, 1007 (1935). 60. H. Kuhn, Pure Appl. Chem. 11, 345 (1965). 61. H. Kuhn, Naturwissenschaften 54, 429 (1967). 62. H. Kuhn and D. Möbius, Angew. Chem. 83, 672 (1971); Int. Ed. Eng. 10, 620 (1971). 63. H. Kuhn, D. Möbius, and H. Bücher, in “Techniques of Chemistry” (A. Weissberger, Ed.), Vol. I, Part IIIb (Physical Methods of Chemistry). Wiley-Interscience, New York, 1972. 64. U. Simon and G. Schön, in “Handbook of Nanostructured Materials and Nanotechnology” (H. S. Nalwa, Ed.), Vol. 3, p. 131. Academic Press, San Diego, 2000. 65. J. H. Fendler, “Membrane Mimetic Chemistry.” Wiley, New York, 1982. 66. A. Ulman, “An Introduction to Ultrathin Organic Films. From Langmuir–Blodgett to Self-Assembly.” Academic Press, Boston, 1991. 67. D. Petridis, P. Falaras, and T. J. Pinnavaia, Inorg. Chem. 31, 3530 (1992). 68. Y. Xiang and G. Villemure, J. Phys. Chem. 100, 7143 (1996). 69. L. A. Vermeulen, Progr. Inorg. Chem. 44, 143 (1997). 70. D. Rong, Y. I. Kim, and T. E. Mallouk, Inorg. Chem. 29, 1531 (1990). 71. G. Cao, H. G. Hong, and T. E. Mallouk, Acc. Chem. Res. 25, 420 (1992). 72. T. E. Mallouk, H.-K. Kim, P. J. Ollivier, and S. W. Keller, in “Comprehensive Supramolecular Chemistry” (G. Alberti and T. Bein, Eds.), Vol. 7, Ch. 6, p. 189. Pergamon, Oxford, 1996. 73. P. Behrens and A. M. Glaue, Monatsh. Chem., in press. 74. P. Behrens, A. M. Glaue, and O. Oellrich, submitted for publication. 75. J. Bauer, P. Behrens, M. Speckbacher, and H. Langhals, submitted to publication. 76. S. Feng and T. Bein, Nature 368, 834 (1994).

Dye/Inorganic Nanocomposites

688 77. K. Hoffmann, F. Marlow, and J. Caro, Adv. Mater. 9, 567 (1997). 78. “Intercalation Chemistry” (M. S. Whittingham and A. J. Jacobson, Eds.). Academic Press, New York, 1982. 79. R. Schöllhorn, in “Inclusion Compounds” (J. L. Atwood, J. E. D. Davies, and D. D. MacNicol, Eds.), p. 249. Academic Press, New York, 1984. 80. A. J. Jacobson, in “Solid-State Chemistry Compounds” (A. K. Cheetham and P. Day, Eds.), Ch. 6, p. 182. Clarendon, Oxford, 1992. 81. D. O’Hare, in “Inorganic Materials” (D. W. Bruce and D. O’Hare, Eds.), Ch. 4, p. 165. Wiley, Chichester, 1992. 82. “Progress in Intercalation Research” (W. Müller-Warmuth and R. Schöllhorn, Eds.). Kluwer, Dordrecht, 1994. 83. A. Lerf, in “Handbook of Nanostructured Materials and Nanotechnology” (H. S. Nalwa, Ed.), Vol. 5, p. 1. Academic Press, San Diego, 2000. 84. S. J. Tsipursky and V. A. Drits, Clay Miner. 19, 177 (1984). 85. J. M. Thomas, in “Intercalation Chemistry” (M. S. Whittingham and A. J. Jacobson, Eds.), p. 55. Academic Press, New York, 1982. 86. K. Jasmund and G. Lagaly, “Tonminerale und Tone.” Steinkopff Verlag, Darmstadt, 1993. 87. “Organo-Clay Complexes and Interactions” (S. Yarif and H. Cross, Eds.). Dekker, New York, 2002. 88. T. Fujita, N. Iyi, T. Kosugi, A. Ando, T. Degushi, and T. Sato, Clays Clay Miner. 45, 77 (1997). 89. R. Sasai, T. Fujita, N. Iyi, H. Itoh, and K. Takagi, Langmuir 18, 6578 (2002). 90. S H. van Olphen, “Clay Colloid Chemistry. For Clay Technologists, Geologists, and Soil Scientists.” Wiley, New York, 1977. 91. G. Lagaly, in “Developments in Ionic Polymers” (A. D. Wilson and J. Prosser, Eds.), Vol. 2, Ch. 2, p. 77. Elsevier, London, 1986. 92. A. J. Jacobson, in “Comprehensive Supramolecular Chemistry” (G. Alberti and T. Bein, Eds.), Vol. 7, Ch. 10, p. 315. Pergamon, Oxford, 1996. 93. M. Ogawa, M. Yamamoto, and K. Kuroda, Clay Miner. 36, 263 (2001). 94. G. Lagaly, Adv. Coll. Interf. Sci. 11, 105 (1979). 95. J. M. Troup and A. Clearfield, Inorg. Chem. 16, 3311 (1977). 96. G. Alberti, in “Comprehensive Supramolecular Chemistry” (G. Alberti and T. Bein, Eds.), Vol. 7, Ch. 5, p. 151. Pergamon, Oxford, 1996. 97. G. Alberti, M. Casciola, U. Costantino, and R. Vivani, Adv. Mater. 8, 291 (1996). 98. A. Clearfield, in “Progress in Intercalation Research” (W. MüllerWarmuth and R. Schöllhorn, Eds.). Kluwer, Dordrecht, 1994. 99. A. Clearfield and U. Costantino, in “Comprehensive Supramolecular Chemistry” (G. Alberti and T. Bein, Eds.), Vol. 7, Ch. 4, p. 107. Pergamon, Oxford, 1996. 100. F. Cavani, F. Trifirò, and A. Vaccari, Catal. Today 11, 173 (1991). 101. F. Trifirò and A. Vaccari, in “Comprehensive Supramolecular Chemistry” (G. Alberti and T. Bein, Eds.), Vol. 7, Ch. 8, p. 251. Pergamon, Oxford, 1996. 102. T. Sato, J. Appl. Chem. 14, 303 (1964). 103. J. P. Thiel, C. K. Chiang, and K. R. Poepelmeier, Chem. Mater. 5, 297 (1993). 104. A. V. Besserguenev, A. M. Fogg, R. J. Francis, S. J. Price, D. O’Hare, V. P. Isupov, and B. P. Tolochko, Chem. Mater. 9, 241 (1997). 105. G. G. Aloisi, N. Coletti, U. Costantino, F. Elisei, L. Latterini, and M. Nocchetti, Langmuir 15, 4454 (1999). 106. G. G. Aloisi, N. Coletti, U. Costantino, F. Elisei, L. Latterini, and M. Nocchetti, Langmuir 16, 10351 (2000). 107. G. G. Aloisi, U. Costantino, F. Elisei, L. Latterini, and M. Nocchetti, Phys. Chem. Chem. Phys. 4, 2792 (2002). 108. G. G. Aloisi, U. Costantino, F. Elisei, L. Latterini, C. Natali, and M. Nocchetti, J. Mater. Chem. 12, 3316 (2002). 109. B. Raveau, Rev. Chim. Miner. 21, 391 (1984).

110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135.

136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146.

B. Raveau, Rev. Inorg. Chem. 9, 37 (1987). S. Andersson and A. D. Wadsley, Acta Cryst. 14, 1245 (1961). M. Gasperin and M. T. Bihan, J. Solid State Chem. 33, 83 (1980). M. Gasperin and M. T. Bihan, J. Solid State Chem. 43, 346 (1982). N. Kinomura, N. Kumada, and F. Muto, J. Chem. Soc. Dalton Trans. 2349 (1985). A. Kudo, A. Tanaka, K. Domen, K.-I. Maruya, K.-I. Aika, and T. Ohishi, J. Catal. 111, 67 (1988). T. Nakato and K. Kuroda, Eur. J. Sol. State Inorg. Chem. 32, 809 (1995). A. Kudo and T. Kondo, J. Mater. Chem. 7, 777 (1997). T. Takata, Y. Furumi, K. Shinohara, A. Tanaka, M. Hara, J. N. Kondo, and K. Domen, Chem. Mater. 9, 1063 (1997). M. M. C. Treacy, S B. Rice, A. J. Jacobson, and J. T. Lewandowski, Chem. Mater. 2, 279 (1990). R. Abe, K. Shinohara, A. Tanaka, M. Hara, J. N. Kondo, and K. Domen, Chem. Mater. 9, 2179 (1997). J. Livage, Chem. Mater. 3, 578 (1991). M. G. Kanatzidis and C. G. Wu, J. Am. Chem. Soc. 111, 4139 (1989). M. G. Kanatzidis, C. G. Wu, H. O. Marcy, D. C. DeGroot, and C. R. Kannewurf, Chem. Mater. 2, 222 (1990). H. Masbah, D. Tinet, M. Crespin, R. Erre, R. Setton, and H. van Damme, J. Chem. Soc. Chem. Commun. 935 (1985). H. Van Damme, M. Letellier, D. Tinet, B. Kihal, and R. Erre, Mater. Res. Bull. 19, 1635 (1984). B. Ackermans, R. A. Schoonheydt, and E. Ruiz-Hitzky, J. Chem. Soc. Faraday Trans. 92, 4479 (1996). T. Kuwahara, H. Tagaya, and J. Kadokawa, Inorg. Chem. Comm. 4, 63 (2001). S. M. Kauzlarich, B. K. Teo, and B. A. Averill, Inorg. Chem. 25, 1209 (1986). J. F. Bringley, J. M. Fabre, and B. A. Averill, J. Amer. Chem. Soc. 112, 4577 (1990). C. Riekel, H. G. Reznik, and R. Schöllhorn, J. Solid State Chem. 34, 253 (1980). A. Hauptmann, A. Lerf, and W. Biberacher, Z. Naturf. B 51, 1571 (1996). ˇ P. Capková, M. Pospíšil, and A. Lerf, Solid State Sci. 4, 671 (2002). A. Schlicht, M. Schwenker, W. Biberacher, and A. Lerf, J. Phys. Chem. B 105, 4867 (2001). S. V. Smith, Chem. Rev. 88, 149 (1988). H. Gies and B. Marler, “Comprehensive Supramolecular Chemistry” (D. D Mac Nicol, F. Toda, and R. Bishop, Eds.), Vol. 6. Pergamon, Oxford, 1996. G. Van de Goor, C. Freyhardt, and P. Behrens, Z. Anorg. Allg. Chem. 621, 311 (1995). W. M. Meier and W. A. Olson, “Atlas of Zeolite Structure Types,” 2nd revised ed. Butterworths, Cambridge, UK, 1987. D. H. Olson, G. T. Kokotallo, S. L. Lawton, and W. M. Meier, J. Phys. Chem. 85, 2238 (1981). E. M. Flanigen, J. M. Bennett, R. W. Grose, J. P. Cohen, R. L. Patton, R. M. Kirchner, and J. V. Smith, Nature 271, 512 (1978). S. T. Wilson, B. M. Lok, C. A. Messina, T. R. Cannon, and E. M. Flanigen, J. Am. Chem. Soc. 104, 1146 (1982). J. M. Bennett, W. J. Dytrych, J. J. Pluth, J. W. Richardson, and J. V. Smith, Zeolites 6, 349 (1986). J. W. Richardson, J. J. Pluth, and J. V. Smith, Acta Cryst. C 43, 1469 (1987). J. W. Richardson, J. J. Pluth, and J. V. Smith, Acta Cryst. B 44, 367 (1988). M. E. Davis, C. Saldarriga, C. Montes, J. Garces, and C. Crowder, Nature 331, 698 (1988). L. B. McCusker, C. Baerlocher, E. Jahn, and M. Bülow, Zeolites 11, 308 (1991). B. M. Lok, C. A. Messina, R. L. Patton, R. T. Gajek, T. R. Cannon, and E. M. Flanigen, J. Am. Chem. Soc. 106, 6092 (1984).

Dye/Inorganic Nanocomposites 147. N. Raji´c, V. Kauˇciˇc, and D. Stojakovi´c, Zeolites 10, 169 (1990). 148. C. M. Cardile, N. J. Trapp, and N. B. Milestone, Zeolites 10, 90 (1990). 149. S. Wohlrab, R. Hoppe, G. Schulz-Ekloff, and D. Wöhrle, Zeolites 12, 862 (1992). 150. R. Hoppe, G. Schulz-Ekloff, D. Wöhrle, E. S. Shpiro, and O. P. Tkachenko, Zeolites 13, 222 (1993). 151. C. T. Cresge, M. E. Leonowicz, W. J. Roth, J. C. Vartuli, and J. S. Beck, Nature 359, 710 (1992). 152. G. Øye, J. Sjöblom, and M. Stöcker, Adv. Coll. Interf. Sci. 89–90, 439 (2001). 153. M. Wark, A. Ortlam, M. Ganschow, G. Schulz-Ekloff, and D. Wöhrle, Ber. Bunsenges. Phys. Chem. 102, 1548 (1998). 154. H. S. Zhou and I. Honma, Adv. Mater. 11, 683 (1999). 155. H. S. Zhou, H. Sasabe, and I. Honma, J. Mater. Chem. 8, 515 (1998). 156. I. Honma and H. S. Zhou, Chem. Mater. 10, 103 (1998). 157. X. He and D. Antonelli, Angew. Chem. 114, 222 (2002); Int. Ed. Engl. 41, 214 (2002). 158. B. F. Jones and E. Galan, in “Hydrous Phyllosilicates” (S. W. Bailey, Ed.), Reviews in Mineralogy, Vol. 19. Miner. Soc. Am., Washington, DC, 1988. 159. C. Detellier et al., submitted for publication 160. H. Dürr, Angew. Chem. 101, 427 (1989); Int. Ed. Engl. 28, 413 (1989). 161. T. J. Marks and M. A. Ratner, Angew. Chem. 107, 167 (1995); Int. Ed. Engl. 34, 155 (1995). 162. N. J. Long, Angew. Chem. 107, 37 (1995); Int. Ed. Engl. 34, 21 (1995). 163. G. Van de Goor, B. Lindlar, J. Felsche, and P. Behrens, J. Chem. Soc. Chem. Commun. 2559 (1995). 164. F. Marlow, G. van de Goor, and P. Behrens, Adv. Mater. 11, 238 (1999). 165. A. Yamagishi, J. Coord. Chem. 16, 131 (1987). 166. M.-H. Ryoo and H. Chon, Microporous Mater. 10, 35 (1997). 167. E. P. Giannelis, D. G. Nocera, and T. J. Pinnavaia, Inorg. Chem. 26, 203 (1987). 168. M. F. Traynor, M. M. Mortland, and T. J. Pinnavaia, Clays Clay Miner. 26, 318 (1978). 169. A. Weiss and G. Roloff, Z. Naturf. B 19, 533 (1964). 170. F. Bergaya and H. Van Damme, Geochim. Cosmochim. Acta 46, 349 (1982). 171. S. S. Cady and T. J. Pinnavaia, Inorg. Chem. 17, 1501 (1978). 172. H. Van Damme, M. Crespin, F. Obrecht, M. I. Cruz, and J. J. Fripiat, J. Coll. Interface Sci. 66, 43 (1978). 173. S. Abdo, M. I. Cruz, and J. J. Fripiat, Clays Clay Miner. 28, 125 (1980). 174. K. A. Carrado and R. E. Winans, Chem. Mater. 2, 328 (1990). 175. V. B. Preobrazhensky, Z. J. Zhernova, A. P. Babichev, I. K. Kobrin, E. A. Lukyanez, and O. L. Kalija, Solid State Comm. 27, 127 (1978). 176. S. Takagi, T. Shimada, T. Yui, and H. Inoue, Chem. Lett. 128 (2001). 177. S. Takagi, T. Shimada, M. Eguchi, T. Yui, H. Yoshida, D. A. Tryk, and H. Inoue, Langmuir 18, 2265 (2002). 178. S. Takagi, D. A. Tryk, and H. Inoue, J. Phys. Chem. B 106, 5455 (2002). 179. B. T. Holland, C. Walkup, and A. Stein, J. Phys. Chem. B 102, 4301 (1998). 180. D. Wöhrle, A. K. Sobbi, O. Franke, and G. Schulz-Ekloff, Zeolites 15, 540 (1995). 181. L. Ukrainczyk, M. Chibwe, T. J. Pinnavaia, and S. A. Boyd, J. Phys. Chem. 98, 2668 (1994). 182. E. P. Giannelis, Chem. Mater. 2, 627 (1990). 183. T. Nakato, Y. Iwata, K. Kuroda, M. Kaneko, and C. Kato, J. Chem. Soc. Dalton Trans. 1405 (1993). 184. I. Y. Park, K. Kuroda, and C. Kato, Chem. Lett. 2057 (1989).

689 185. K. A. Carrado, P. Thiyagarajan, R. E. Winans, and R. E. Botto, Inorg. Chem. 30, 794 (1991). 186. M. Ehrl, F. W. Deeg, C. Bräuchle, A. Sobbi, G. Schulz-Ekloff, and D. Wöhrle, J. Phys. Chem. 98, 47 (1994). 187. J. M. Krüger and H. D. Brauer, Ber. Bunsenges. Phys. Chem. 102, 1554 (1998). 188. T. Itoh, K. Yano, Y. Inada, and Y. Fukushima, J. Mater. Chem. 12, 3304 (2002). 189. S. Bonnet, C. Forano, A. de Roy, J. P. Besse, P. Maillard, and M. Momentau, Chem. Mater. 8, 1962 (1996). 190. K. A. Carrado and S. R. Wassermann, Chem. Mater. 8, 219 (1996). 191. Z. Chernia and D. Gill, Langmuir 15, 1625 (1999). 192. F. B. Abdelrazzaq, R. C Kwong, and M. E. Thompson, J. Am. Chem. Soc. 124, 4796 (2002). 193. K. A. Carrado, J. E. Forman, R. E. Botto, and R. E. Winans, Chem. Mater. 5, 472 (1993). 194. M. R. Wasielewski, Chem. Rev. 92, 435 (1992). 195. C. L. Bird and A. T. Kuhn, Chem. Soc. Rev. 10, 49 (1981). 196. H. J. Schäfer, Top. Current Chem. 142, 101 (1987). 197. L. Michaelis and E. S. Hill, J. Gen. Physiol. 16, 859 (1933). 198. G. Villemure, H. Kodoma, and C. Detellier, Can. J. Chem. 63, 1139 (1985). 199. J. B. Weber and D. C. Scott, Science 152, 1400 (1966). 200. M. Raupach, W. W. Emerson, and P. G. Slade, J. Coll. Interface Sci. 69, 398 (1979). 201. T. Endo, T. Sato, and M. Shimada, J. Phys. Chem. Solids 47, 799 (1986). 202. L. Margulies, H. Rozen, and S. Nir, Clays Clay Miner. 36, 270 (1988). 203. R. Cohen and S. Yariv, J. Chem. Soc. Faraday Trans. I 80, 1705 (1984). 204. M. Bockstette, D. Wöhrle, I. Braun, and G. Schulz-Ekloff, Microporous Mesoporous Mater. 23, 83 (1998). 205. S. K. Lam and D. Lo, Chem. Phys. Lett. 281, 35 (1997). 206. M. J. T. Estévez, F. L. Arbeloa, T. L. Arbeloa, and I. L. Arbeloa, J. Colloid Interface Sci. 171, 439 (1995). 207. M. J. T. Estévez, F. L. Arbeloa, T. L. Arbeloa, and I. L. Arbeloa, J. Colloid Interface Sci. 162, 412 (1994). 208. F. L. Arbeloa, R. Chaudhuri, T. L. Arbeloa, and I. L. Arbeloa, J. Colloid Interface Sci. 246, 281 (2002). 209. H. Van Damme, M. Crespin, M. I. Cruz, and J. J. Fripiat, Clays Clay Miner. 25, 19 (1977). 210. J. Bujdák and N. Iyi, Clays Clay Miner. 50, 446 (2002). 211. G. Calzaferri and N. Gfeller, J. Phys. Chem. 96, 3428 (1992). 212. V. Ramamurthy, D. D. Sanderson, and D. F. Eaton, J. Am. Chem. Soc. 115, 10438 (1993). 213. M. Ehrl, H. W. Kindervater, F. W. Deeg, and C. Bräuchle, J. Phys. Chem. 98, 11756 (1994). 214. D. Brühwiler, N. Gfeller, and G. Calzaferri, J. Phys. Chem. B 102, 2923 (1998). 215. N. Gfeller, S. Megelski, and G. Calzaferri, J. Phys. Chem. B 102, 2433 (1998). 216. T. Endo, N. Nakada, T. Sato, and M. Shimada, J. Phys. Chem. Solids 49, 1423 (1988). 217. F. Marlow, M. D. McGehee, D. Zhao, B. F. Chmelka, and G. D. Stucky, Adv. Mater. 11, 632 (1999). 218. U. Narang, R. Wang, P. N. Prasad, and F. V. Bright, J. Phys. Chem. 98, 17 (1994). 219. D. Levy and D. Avnir, J. Photochem. Photobiol. A: Chem. 57, 41 (1991). 220. A. Dubois, M. Canva, A. Brun, F. Chaput, and J.-P. Boilot Synth. Met. 81, 305 (1996). 221. G. Wirnsberger and G. D. Stucky, Chem. Mater. 12, 2525 (2000). 222. F. Del Monte and D. Levy, J. Phys. Chem. B 103, 8080 (1999). 223. J. L. Meinershagen and T. Bein, J. Am. Chem. Soc. 121, 448 (1999). 224. R. Hoppe, G. Schulz-Ekloff, D. Wöhrle, C. Kirschock, H. Fuess, L. Uyterhoeven, and R. A. Schoonheydt, Adv. Mater. 7, 61 (1995).

Dye/Inorganic Nanocomposites

690 225. 226. 227. 228. 229. 230. 231. 232. 233. 234. 235. 236. 237. 238. 239. 240. 241. 242. 243. 244. 245. 246. 247. 248. 249. 250. 251. 252. 253. 254. 255. 256. 257. 258. 259. 260. 261. 262. 263. 264. 265. 266.

K. Bergman and C. T. O’Konski, J. Phys. Chem. 67, 2169 (1963). P. T. Hang and G. W. Brindley, Clays Clay Miner. 18, 203 (1970). J. Cenens and R. A. Schoonheydt, Clay Miner. 23, 205 (1988). J. Cenens and R. A. Schoonheydt, Clays Clay Miner. 36, 214 (1988). R. A. Schoonheydt and L. Heughebaert, Clay Miner. 27, 91 (1992). F. Gessner, C. Schmitt, and M. G. Neumann, Langmuir 10, 3749 (1994). J. Bujdák and P. Komadel, J. Chem. Phys. B 101, 9065 (1997). A. P. P. Cione, M. G. Neumann, and F. Gessner, J. Coll. Interf. Sci. 198, 106 (1998). K. Y. Jacobs and R. A. Schoonheydt, J. Coll. Interf. Sci. 220, 103 (1999). J. Bujdák, M. Janek, J. Madejovà, and P. Komadel, Clays Clay Miner. 49, 244 (2001). J. Bujdák, N. Iyi, and T. Fujita, Clay Miner. 37, 121 (2002). D. Garfinkel-Shweky and S. Yariv, J. Coll. Interf. Sci. 188, 168 (1997). D. Garfinkel-Shweky and S. Yariv, Clay Miner. 34, 459 (1999). L. Michaelis and S. Granick, J. Am. Chem. Soc. 63, 1636 (1941). K. Schwabe and H. Berg, Z. Phys. Chem. 204, 78 (1955). S. H. de Araujo-Nicolai, P. R. P. Rodrigues, S. M. L. Agostinho, and L. C. Rubin, J. Electroanal. Chem. 527, 103 (2002). S. Ameta, S. Khamesra, A. K. Cittora, and K. M. Gangotri, Int. J. Energ. Res. 13, 643 (1989). D. F. Duxbury, Chem. Rev. 93, 381 (1993). R. Hoppe, G. Alberti, U. Costantino, C. Diongi, G. Schulz-Ekloff, and R. Vivani, Langmuir 13, 7252 (1997). D. Avnir, Z. Grauer, S. Yariv, D. Huppert, and D. Rojanski, Nouv. J. Chim. 10, 153 (1986). L. Margulies, H. Rozen, and E. Cohen, Clays Clay Miner. 36, 159 (1988). D. N. Simon, R. Czolk, and H. J. Ache, Thin Solid Films 260, 107 (1995). Z. Chernia, D. Gill, and S. Yariv, Langmuir 10, 3988 (1994). K. Yamaoka and R. Sasai, J. Coll. Interf. Sci. 225, 82 (2000). P. G. Lacroix, R. Clement, K. Nakatani, J. Zyss, and I. Ledoux, Science 263, 658 (1994). I. Lagadic, P. G. Lacroix, and R. Clement, Chem. Mater. 9, 2004 (1997). D. Gonbeau, T. Coradin, and R. Clement, J. Phys. Chem. B 103, 3545 (1999). J. Kim, M. Lee, J. H. Yang, and J. H. Choy, J. Phys. Chem. A 104, 1388 (2000). K. Izawa, N. Okamoto, and O. Sugihara, Jpn. J. Appl. Phys. 32, 807 (1993). C. Schomburg, D. Wöhrle, and G. Schulz-Ekloff, Zeolites 17, 232 (1996). K. Hoffmann, U. Resch-Genger, and F. Marlow, Microporous Mesoporous Mater. 41, 99 (2000). M. Ogawa, Chem. Mater. 8, 1347 (1996). M. Ogawa and A. Ishikawa, J. Mater. Chem. 8, 463 (1998). M. Ogawa, J. Mater. Chem. 12, 3304 (2002). S. K. Doughty, G. J. Simpson, and K. L. Rowlen, J. Am. Chem. Soc. 120, 7997 (1998). H. E. Katz, W. L. Wilson, and G. Scheller, J. Am. Chem. Soc. 116, 6636 (1994). E. C. Y. Yan and K. B. Eisenthal, J. Phys. Chem. B 103, 6056 (1999). I. Y. Park, K. Kuroda, and C. Kato, J. Chem. Soc. Dalton Trans. 3071 (1990). C. V. Kumar, A. Chaudhari, and G. L. Rosenthal, J. Am. Chem. Soc. 116, 403 (1994). D. W. Murphy and G. W. Hull, J. Chem. Phys. 62, 967 (1975). N. Miyamoto, K. Kuroda, and M. Ogawa, J. Am. Chem. Soc. 123, 6949 (2001). A. Corma, V. Fornés, H. García, M. A. Miranda, J. Primo, and M.-J. Sabater, J. Am. Chem. Soc. 116, 2276 (1994).

267. T. Endo, N. Nakada, T. Sato, and M. Shimada, J. Phys. Chem. Solids 50, 133 (1989). 268. S. D. Cox, T. E. Gier, and G. D. Stucky, Chem. Mater. 2, 609 (1990). 269. F. Marlow, M. Wübbenhorst, and J. Caro, J. Phys. Chem. 98, 12315 (1994). 270. F. Marlow, J. Caro, L. Werner, J. Kornatowski, and S. Dähne, J. Phys. Chem. 9, 11286 (1993). 271. K. Hoffmann, F. Marlow, J. Caro, and S. Dähne, Zeolites 16, 138 (1996). 272. W. K. Kuk and Y.-D. Huh, J. Mater. Chem. 7, 1933 (1997). 273. S. Miyata, Clays Clay Miner. 31, 305 (1983). 274. N. Khaorapapong, K. Kuroda, and M. Ogawa, Clays Clay Miner. 50, 428 (2002). 275. B. D. Gupta and D. K. Sharma, Optics Commun. 140, 32 (1997). 276. C. Schomburg, M. Wark, Y. Rohlfing, G. Schulz-Ekloff, and D. Wöhrle, J. Mater. Chem. 11, 2014 (2001). 277. G. Wirnsberger, B. J. Scott, B. F. Chmelka, and G. D. Stucky, Adv. Mater. 12, 1450 (2000). 278. H. Tomioka and T. Itoh, J. Chem. Soc. Chem. Commun. 532 (1991). 279. T. Seki and K. Ichimura, Macromolecules 23, 31 (1990). 280. K. Takagi, T. Kurematsu, and Y. Sawaki, J. Chem. Soc. Perkin Trans. 2 1517 (1991). 281. N. Röhrlich, E. Löffler, B. Zibrowius, L. Chassot, and W. F. Hölderich, J. Chem. Soc. Fararday Trans. 94, 609 (1998). 282. L.-T. Yuen, J. S. Geilfuss, and S. I. Zones, Microporous Mater. 12, 229 (1997). 283. R. Hoppe, G. Schulz-Ekloff, D. Wöhrle, C. Kirschhock, and H. Fuess, Langmuir 10, 1517 (1994). 284. H. Van Damme, F. Obrecht, and M. Letellier, Nouv. J. Chim. 8, 681 (1984). 285. P. Lacroix, J. P. Audière, and R. Clement, J. Chem. Soc. Chem. Comm. 536 (1989). 286. L. Lomas, P. Lacroix, J. P. Audière, and R. Clement, J. Mater. Chem. 1, 475 (1991). 287. T. E. Sutto, B. A. Averill, and J. M. Fabre, Synth. Met. 55–57, 2019 (1993). 288. S. M. Kauzlarich, J. L. Stanton, J. Faber, and B. A. Averill, J. Am. Chem. Soc. 108, 7946 (1986). 289. R. Backov, B. Bonnet, D. J. Jones, and J. Rozière, Chem. Mater. 9, 1812 (1997). 290. R. A. DellaGuardi and J. K. Thomas, J. Phys. Chem. 88, 964 (1984). 291. P. Labbé and G. Reverdy, Langmuir 4, 419 (1988). 292. X. Liu and J. K. Thomas, Langmuir 7, 2808 (1991). 293. R. J. Kavanagh, K. K. Iu, and J. K. Thomas, Langmuir 8, 3008 (1992). 294. X. Liu, K. K. Iu, and J. K. Thomas, Langmuir 8, 539 (1992). 295. B. Brahimi, P. Labbé, and G. Reverdy, Langmuir 8, 1908 (1992). 296. M. Ogawa, T. Aono, K. Kuroda, and C. Kato, Langmuir 9, 1529 (1993). 297. K. B. Yoon, T. J. Huh, D. R. Corbin, and J. K. Kochi, J. Phys. Chem. 97, 6492 (1993). 298. G. Berkovic, V. Krongauz, and V. Weiss, Chem. Rev. 100, 1741 (2000). 299. W. F. Bradley, R. E. Grim, and G. L. Clark, Z. Kristallogr. 97, 216 (1937). 300. D. Krenske, S. Abdo, H. Van Damme, M. Cruz, and J. J. Fripiat, J. Phys. Chem. 84, 2447 (1980). 301. G. Decher, in “Comprehensive Supramolecular Chemistry” (J. P. Sauvage and M. W. Hosseini, Eds.), Vol. 9, Ch. 14, p. 507. Pergamon, Oxford, 1996. 302. R. K. Iler, J. Coll. Interf. Sci. 21, 569 (1966). 303. J. Schmitt, T. Grünewald, G. Decher, P. S. Persham, K. Kjaer, and M. Lösche, Macromolecules 26, 7058 (1993). 304. Y. Lvov, K. Ariga, I. Ichinose, and T. Kunitake, J. Amer. Chem. Soc. 117, 6117 (1995).

Dye/Inorganic Nanocomposites 305. S. W. Keller, H.-N. Kim, and T. E. J. Mallouk, Amer. Chem. Soc. 116, 8817 (1994). 306. G. S. Ferguson and E. R. Kleinfeld, Adv. Mater. 7, 414 (1995). 307. J. H. Fendler and F. C. Meldrum, Adv. Mater. 7, 607 (1995). 308. J. H. Fendler, Chem. Mater. 8, 1616 (1996). 309. N. A. Kotov, T. Haraszti, L. Turi, G. Zawala, R. E. Geer, I. Dékány, and J. Fendler, J. Am. Chem. Soc. 119, 6821 (1997). 310. B. van Duffel, R. Schoonheydt, C. P. M. Grim, and F. C. DeSchryver, Langmuir 15, 7520 (1999). 311. B. van Duffel, T. Verbiest, S. Van Elshocht, A. Persoons, F. C. DeSchryver, and R. Schoonheydt, Langmuir 17, 1243 (2001). 312. H. E. Katz, G. Scheller, T. M. Putvinski, M. L. Schilling, W. L. Wilson, and C. E. D. Chidsey, Science 254, 1485 (1991). 313. S. B. Ungashe, W. L. Wilson, H. E. Katz, G. Scheller, and T. M. Putvinski, J. Am. Chem. Soc. 114, 8717 (1992). 314. L. A. Vermeulen and M. E. Thompson, Nature 358, 656 (1992). 315. Y. Lvov, F. Essler, and G. Decher, J. Phys. Chem. 97, 13773 (1993). 316. K. Inukai, Y. Hotta, M. Taniguchi, S. Tomura, and A. Yamagishi, J. Chem. Soc. Chem. Commun. 959 (1994). 317. N. Kotov, A. Meldrum, F. C. Fendler, J. H. Tombácz, and I. Dekany, Langmuir 10, 3797 (1994). 318. Y. Umemura, A. Yamagishi, R. Schoonheydt, A. Persoons, and F. DeSchryver, Langmuir 17, 449 (2001). 319. K. Tamura, H. Setsuda, M. Taniguchi, and A. Yamagishi, Langmuir 15, 6915 (1999). 320. A. Tavolaro and E. Driolo, Adv. Mater. 11, 975 (1999). 321. S. Feng and T. Bein, Science 265, 1839 (1994). 322. J. Caro, G. Finger, J. Kornatowski, J. Richter-Mendau, L. Werner, and B. Zibrowius, Adv. Mater. 4, 273 (1992). 323. J. M. Lehn, Angew. Chem. Int. Ed. Engl. 29, 1304 (1990). 324. A. Nangia and G. R. Desiraju, Acta Cryst. A 54, 934 (1998). 325. A. Gavezzotti, J. Am. Chem. Soc. 113, 4622 (1991). 326. Cerius2 documentation Molecular Simulations Inc. (CD-ROM), San Diego, June 2000. 327. P. Comba and T. W. Hambley, “Molecular Modeling of Inorganic Compounds.” VCH, Weinheim, 1995. 328. D. Frenkel and B. Smit, “Understanding Molecular Simulation.” Academic Press, San Diego, 1996. ˇ 329. P. Capková, J. V. Burda, Z. Weiss, and H. Schenk, J. Mol. Model. 5, 8 (1999). 330. A. K. Rappé, C. J. Casewit, K. S. Colwell, W. A. Goddard III, and W. M. Skiff, J. Am. Chem. Soc. 114, 10024 (1992). 331. P. Dauber-Osguthorpe, V. A. Roberts, D. J. Osguthorpe, J. Wolff, M. Genest, and A. T. Hagler, Proteins: Structure, Function Genetics 4, 31 (1988). 332. M. Clark, R. D. Cramer III, and N. Van Opdenbosh, J. Comp. Chem. 10, 982 (1989). 333. H. Sun, J. Comp. Chem. 15, 752 (1994). 334. S. L. Mayo, B. D. Olafson, and W. A. Goddard III, J. Phys. Chem. 94, 8897 (1990). 335. A. K. Rappé and W. A. Goddard III, J. Phys. Chem. 95, 3358 (1991). 336. P. P. Ewald, Ann. Phys. Leipzig 64, 253 (1921). 337. N. Karasawa and W. A. Goddard III, J. Phys. Chem. 93, 7320 (1989). ˇ 338. P. Koudelka and P. Capková, J. Mol. Model 8, 184 (2002). 339. P. Erk, in “Crystal Engineering: From Molecules and Crystals to Materials” (D. Braga and G. Orpen, Eds.), p. 143. Kluwer Academic, Dordrecht, 1999. 340. P. Verwer and F. J. J. Leusen, Rev. Comput. Chem. 12, 327 (1998). 341. K. D. M. Harris, B. M. Kariuki, and R. L. Johnston, in “Advances in Structure Analysis” (R. Kužel and J. Hašek, Eds.), p. 190. Czech and Slovak Crystallographic Association, Prague, 2001. 342. W. A. Dollase, J. Appl. Cryst. 19, 267 (1986). 343. P. Suortti, J. Appl. Cryst. 5, 325 (1972). 344. F. R. Gamble, F. J. DiSalvo, R. A. Klemm, and T. H. Geballe, Science 168, 568 (1970).

691 345. R. H. Friend and A. D. Yoffe, Adv. Phys. 36, 1 (1987). 346. A. Schlicht, Thesis, TU München, 1999. 347. A. Meetsma, G. A. Wiegers, R. J. Haange, and J. L. De Boer, Acta Cryst. C 46, 1598 (1990). 348. A. Kahn-Harari, R. E. Ballard, and E. K. Norris, Acta Cryst. B 29, 1124 (1973). 349. J. Breu, N. Raj, and C. R. A. Catlow, J. Chem. Soc. Dalton Trans. 6, 835 (1999). 350. J. Méring and A. Oberlin, Clays Clay Miner. 27, 3 (1967). 351. M. Pospíšil, Complex Structure Analysis of Intercalates Using Molecular Simulations, Thesis, Charles University, Prague, 2002. ˇ 352. M. Pospíšil, P. Capková, H. Weissmannová, Z. Klika, M. Trchová, M. Chmielová, and Z. Weiss, J. Mol. Model, in press. 353. R. Chaudhuri, F. L. Arbeloa, and I. L. Arbeloa, Langmuir 16, 1285 (2000). 354. A. Vahedi-Faridi and S. Guggenheim, Clay Clay Miner. 45, 859 (1997). 355. A. Vahedi-Faridi and S. Guggenheim, Clay Clay Miner. 47, 338 (1999). 356. P. G. Slade and P. A. Stone, Clay Clay Miner. 32, 223 (1984). 357. J. Méring, in “Soil Components” (J. E. Gieseking, Ed.), Vol. 2, p. 110. Springer-Verlag, Berlin, 1975. ˇ 358. P. Capková, J. V. Burda, Z. Weiss, and H. Schenk, J. Mol. Model. 5, 8 (1999). 359. P. G. Slade, C. Dean, P. K. Schultz, and P. G. Self, Clays Clay Miner. 35, 177 (1987). 360. P. G. Slade and M. Raupach, Clays Clay Miner. 30, 297 (1982). 361. G. Villemure, Clays Clay Miner. 38, 623 (1990). 362. J. Breu and C. R. A. Catlow, Inorg. Chem. 34, 4504 (1995). 363. V. Joshi, D. Koktar, and P. K. Gosh, Proc. Indian Acad. Sci: Chem. Sci. 102, 203 (1990). 364. R. A. Schoonheydt, P. Depauw, D. Vliers, and F. C. DeSchryver, J. Phys. Chem. 88, 5113 (1984). 365. P. V. Kamat, K. R. Gopidas, T. Mukherjee, V. Joshi, D. Koktar, V. S. Pathak, and P. K. Gosh, J. Phys. Chem. 95, 10009 (1991). 366. N. Shimizu, S. Hashimoto, T. Takemura, M. Kawasaki, and A. Yamagishi, Clays controlling the environment, in “Proc. 10th International Clay Conference” (G. J. Churchman, R. W. Fitzpatrick, and R. A. Eggleton, Eds.). CSIRO, Melbourne, 1995. 367. H. Van Koningsveld and J. H. Koegler, Microporous Mater. 9, 71 (1997). 368. L. Werner, J. Caro, G. Finger, and J. Kornatowski, Zeolites 12, 658 (1992). 369. G. Reck, F. Marlow, J. Kornatowski, W. Hill, and J. Caro, J. Phys. Chem. 100, 1698 (1996). 370. B. Hennesy, S. Megelski, C. Marcolli, V. Shklover, C. Bärlocher, and G. Calzaferri, J. Phys. Chem. B 103, 3340 (1999). 371. E. H. Megehee, C. E. Johnson, and R. Eisenberg, Inorg. Chem. 28, 2423 (1989). 372. K. Nakamura, Y. Kai, N. Yasuoka, and N. Kasai, Bull. Chem. Soc. Jpn. 54, 3300 (1981). 373. H. Wolkers, R. Stegemann, G. Frenking, K. Dehnicke, D. Fenzke, and G. Baum, Z. Naturf. 48b, 1341 (1993). 374. M. Alvaro, G. A. Facey, H. García, S. García, and J. C. Scaiano, J. Phys. Chem. 100, 18173 (1996). 375. P. K. Ghosh and A. J. Bard, J. Phys. Chem. 88, 5519 (1984). 376. H. Nijs, M. I. Cruz, J. J. Fripiat, and H. van Damme, Nuov. J. Chim. 6, 551 (1982). 377. N. J. Turro, C. V. Kumar, Z. Grauer, and J. K. Barton, Langmuir 3, 1056 (1987). 378. D. P. Vliers, D. Collin, R. A. Schoonheydt, and F. C. DeSchryver, Langmuir 2, 165 (1986). 379. D. P. Vliers, R. A. Schoonheydt, and F. C. DeSchryver, J. Chem. Soc. Faraday Trans. 81, 2009 (1985). 380. T. Nakamura and J. K. Thomas, Langmuir 1, 568 (1985). 381. W. De Wilde, G. Peeters, and J. H. Lundsford, J. Phys. Chem. 84, 2306 (1980).

Dye/Inorganic Nanocomposites

692 382. 383. 384. 385. 386.

387. 388. 389. 390. 391. 392. 393. 394. 395. 396. 397. 398. 399. 400. 401. 402. 403. 404. 405. 406. 407. 408. 409. 410.

411. 412. 413. 414. 415. 416. 417. 418. 419. 420. 421. 422.

P. K. Dutta and J. A. Incavo, J. Phys. Chem. 91, 4443 (1987). P. K. Dutta and W. Turbeville, J. Phys. Chem. 96, 9410 (1992). M. Borja and P. K. Dutta, Nature 362, 43 (1993). J. S. Krueger, J. E. Mayer, and T. E. Mallouk, J. Am. Chem. Soc. 110, 8332 (1988). E. H. Yonemoto, Y. I. Kim, R. H. Schmehl, J. O. Wallin, B. A. Shoulders, B. K. Richardson. J. F. Haw, and T. E. Mallouk, J. Am. Chem. Soc. 116, 10557 (1994). K. F. Mongey, J. G. Vos, B. D. MacCraith, and C. M. McDonagh, Coord. Chem. Rev. 185–186, 417 (1999). K. Sakoda and K. Kominami, Chem. Phys. Lett. 216, 270 (1993). L. Persaud, A. J. Bard, A. Campion, M. A. Fox, T. E. Mallouk, S. E. Webber, and J. M. White, J. Am. Chem. Soc. 109, 7309 (1987). Y. Yang, P. Wang, and G. Zhou, Anal. Bioanal. Chem. 372, 712 (2002). H. Miyata, Y. Sugahara, K. Kuroda, and C. Kato, J. Chem. Soc. Faraday Trans. 1 83, 1851 (1987). T. Nakato, K. Kuroda, and C. Kato, Chem. Mater. 4, 128 (1992). H. Miyata, Y. Sugahara, K. Kuroda, and C. Kato, J. Chem. Soc. Faraday Trans. 1 84, 2677 (1988). A. Bose, P. He, C. Liu, B. D. Ellman, R. J. Twieg, and S. D. Huang, J. Am. Chem. Soc. 124, 4 (2002). Z. Li, C. M. Wang, L. Persaud, and T. E. Mallouk, J. Phys. Chem. 2592 (1988). G. Villemure, C. Detellier, and A. G. Szabo, J. Am. Chem. Soc. 108, 4658 (1986). G. Villemure, C. Detellier, and A. G. Szabo, Langmuir 7, 1215 (1991). O. Poizot, C. Sourisseau, and Y. Mathey, J. Chem. Soc. Faraday Trans. I 80, 3257 (1984). O. Poizot, C. Sourisseau, and C. Giannotti, Chem. Phys. Lett. 122, 129 (1984). O. Poizot, C. Sourisseau, and J. Corset, J. Mol. Struct. 143, 203 (1986). G. Calzaferri, M. Lanz, and J. W. Li, J. Chem. Soc. Chem. Commun. 1313 (1995). M. Alvaro, H. García, S. García, F. Márquez, and J. C. Scaiano, J. Phys. Chem. B 101, 3043 (1997). F. Marlow and K. Hoffmann, Ber. Bunsenges. Phys. Chem. 101, 1731 (1997). D. Levy and D. Avnir, J. Phys. Chem. 92, 4734 (1988). D. Preston, J. C. Pouxviel, T. Novinson, W. C. Kasca, B. Dunn, and J. I. Zink, J. Phys. Chem. 94, 4167 (1990). J. Biteau, F. Chaput, and J.-P. Boilot, J. Phys. Chem. 100, 9024 (1996). B. Schaudel, C. Guermeur, C. Sanchez, K. Nakatani, and J. A. Delaire, J. Mater. Chem. 7, 61 (1997). H. Tagaya, T. Nagaoka, T. Kuwahara, M. Karasu, J.-I. Kadokawa, and K. Chiba, Microporous Mesoporous Mater. 21, 395 (1998). D. Levy, Chem. Mater. 9, 2666 (1997). G. Calzaferri, D. Brühwiler, S. Megelski, M. Pfenninger, M. Pauchard, B. Hennesy, H. Maas, A. Devaux, and U. Graf, Solid State Sci. 2, 421 (2000). G. Calzaferri, M. Pauchard, H. Maas, S. Huber, A. Khatyr, and T. Schaafsma, J. Mater. Chem. 12, 1 (2002). H. García and H. D. Roth, Chem. Rev. 102, 3947 (2002). G. Scheibe, Angew. Chem. 49, 563 (1936). L. Antonov, G. Gergov, V. Petrov. M. Kubista, and J. Nygren, Talanta 49, 99 (1999). H. Kuhn, Coll. Surf. A 171, 3 (2000). G. Scheibe, Angew. Chem. 50, 51 (1937). E. E. Jelley, Nature 138, 1009 (1936). E. E. Jelley, Nature 139, 631 (1937). D. Möbius, Adv. Mater. 7, 437 (1995). V. G. Kuykendall and J. K. Thomas, Langmuir 6, 1350 (1990). C. Breen and B. Rock, Clay Miner. 29, 179 (1994). G. J. Kavarnos and N. J. Turro, Chem. Rev. 86, 401 (1986).

423. C.-H. Yu, S. Q. Newton, M. A. Norman, D. M. Miller, L. Schäfer, and B. J. Teppen, Clays Clay Miner. 48, 665 (2000). 424. M. J. Avena, L. E. Valenti, V. Pfaffen, and C. P. De Pauli, Clays Clay Miner. 49, 168 (2001). 425. Y. Xiang, G. Villemure, and C. Detellier, Clays Clay Miner. 40, 362 (1992). 426. R. Hoppe, G. Schulz-Ekloff, D. Wöhrle, M. Ehrl, and C. Bräuchle, in “Zeolite Chemistry and Catalysis” (P. A. Jacobs, N. I. Jaeger, L. Kubelková, and B. Wichterlová, Eds.), p. 199. Elsevier, Amsterdam, 1991. 427. W. J. MacNeal and J. A. Killian, J. Am. Chem. Soc. 48, 740 (1926). 428. C. B. Sunwar and H. Bose, J. Coll. Interf. Sci. 136, 54 (1990). 429. T. Coradin, R. Clement, P. G. Lacroix, and K. Nakatani, Chem. Mater. 8, 2153 (1996). 430. A. Habti, D. Keravis, P. Levitz, and H. van Damme, J. Chem. Soc. Faraday Trans. 2 80, 67 (1984). 431. M. Ogawa, M. Inagaki, N. Kodoma, K. Kuroda, and C. Kato, J. Phys. Chem. 97, 3819 (1993). 432. N. Gfeller, S. Megelski, and G. Calzaferri, J. Phys. Chem. B 103, 1250 (1999). 433. M. Pauchard, A. Devaux, and G. Calzaferri, Chem. Eur. J. 6, 3456 (2000). 434. G. Renger, in “Concepts in Photobiology: Photosynthesis and Photo-morphogenesis” (G. S. Singhal, G. Renger, S. K. Sopory, K.-D. Irgang, and Govindjee, Eds.). Narosa, New Dehli, 1999. 435. D. Gust, T. A. Moore, and A. L. Moore, Acc. Chem. Res. 34, 40 (2001). 436. D. G. Whitten, Acc. Chem. Res. 13, 83 (1980). 437. C. R. Bock, T. J. Meyer, and D. G. Whitten, J. Am. Chem. Soc. 96, 4710 (1974). 438. P. Keller, A. Moradpour, E. Armouyal, and H. B. Kagan, Nouv. J. Chim. 4, 377 (1980). 439. J. L. Snover and M. E. Thompson, J. Am. Chem. Soc. 116, 765 (1994). 440. G. Sprintschnik, H. W. Sprintschnik, P. P. Kirsch, and D. G. Whitten, J. Am. Chem. Soc. 98, 2337 (1976). 441. Y. I. Kim, S. Salim, M. J. Huq, and T. E. Mallouk, J. Am. Chem. Soc. 113, 9561 (1991). 442. M. Kirch, J.-M. Lehn, and J.-P. Sauvage, Helv. Chim. Acta 62, 1345 (1979). 443. V. Balzani, A. Juris, M. Venturi, S. Campagna, and S. Serroni, Chem. Rev. 96, 759 (1996). 444. L. Sun, L. Hammarström, B. Åkermark, and S. Styring, Chem. Soc. Rev. 30, 36 (2001). 445. R. Lomoth, T. Häupl, O. Johansson, and L. Hammarström, Chem. Eur. J. 8, 102 (2001). 446. E. H. Yonemoto, G. B. Saupe, R. H. Schmehl, S. M. Hubig, R. R. Riley, B. L. Iverson, and T. E. Mallouk, J. Am. Chem. Soc. 116, 4786 (1994). 447. J. R. Kincaid, Chem. Eur. J. 6, 4055 (2000). 448. M. Sykora and J. R. Kincaid, Nature 387, 162 (1997). 449. L. A. Vermeulen, J. L. Snover, L. S. Sapochak, and M. E. Thompson, J. Am. Chem. Soc. 115, 11767 (1993). 450. D. M. Kaschak, S. A. Johnson, C. C. Waraksa, J. Pogue, and T. E. Mallouk, Coord. Chem. Rev. 185–186, 403 (1999). 451. W. H. Quayle and J. H. Lunsford, Inorg. Chem. 21, 97 (1982). 452. T. Sekine, J. Yoshimura, A. Tanaka, K. Domen, K.-I. Maruya, and T. Onishi, Bull. Chem. Soc. Jpn. 63, 2107 (1990). 453. K. Sayama, A. Tanaka, K. Domen, K. Maruya, and T. Onishi, J. Phys. Chem. 95, 1345 (1991). 454. A. Kudo, K. Sayama, A. Tanaka, K. Asakura, K. Domen, K. Maruya, and T. Onishi, J. Catal. 120, 337 (1989). 455. J.-M. Lehn, J.-P. Sauvage, and R. Ziessel, Nouv. J. Chim. 4, 623 (1980). 456. Y. Kim. S. J. Atherton, E. S. Brigham, and T. E. Mallouk, J. Phys. Chem. 97, 11802 (1993). 457. H. Maas and G. Calcaferri, Angew. Chem. 114, 2389 (2002).

Dye/Inorganic Nanocomposites 458. Special issue of Chemical Reviews devoted exclusively to “Photochromism: Memories and Switches” (M. Irie, guest editor), Chem. Rev. 100(5) (2000). 459. M. Ueda, H.-B. Kim, T. Ikeda, and K. Ishimura, Chem. Mater. 4, 1229 (1992). 460. M. Ueda, H.-B. Kim, and K. Ishimura, Chem. Mater. 6, 1771 (1994). 461. D. J. Williams, Angew. Chem. Int. Ed. Engl. 23, 690 (1984). 462. S. Cooper and P. K. Dutta, J. Phys. Chem. 94, 114 (1990). 463. T. M. Putvinski, M. L. Schilling, H. E. Katz, C. E. D. Chidsey, A. M. Mujsce, and A. B. Emerson, Langmuir 6, 1567 (1990). 464. M. Ogawa, M. Takahashi, and K. Kuroda, Chem. Mater. 6, 715 (1992). 465. J. Breu, A. Stoll, K. G. Lange, and T. Probst, Phys. Chem. Chem. Phys. 3, 1232 (2001).

693 466. J. Breu, W. Seidl, A. Stoll, K. G. Lange, and T. U. Probst, Chem. Mater. 13, 4213 (2001). 467. J.-M. Lehn, “Supramolecular Chemistry.” VCH, Weinheim, 1995. 468. J. C. Scaiano and H. García, Acc. Chem. Res. 32, 783 (1999). 469. R. T. Hayes, M. R. Wasielewski, and D. Gosztola, J. Amer. Chem. Soc. 122, 5563 (2000). 470. A. S. Lukas, S. E. Miller, and M. R. Wasielewski, J. Phys. Chem. B 104, 931 (2000). 471. H. Imahori, H. Norieda, H. Yamada, Y. Nishimura, I. Yamazaki, Y. Sakata, and S. Fukuzumi, J. Amer. Chem. Soc. 123, 100 (2001). 472. W. L. Jido, T. Lee, and T. Pinnavaia, Adv. Mater. 8, 79 (1996). 473. W. L. Jido and T. Pinnavaia, J. Solid State Chem. 139, 281 (1998). 474. W. L. Jido and T. Pinnavaia, Chem. Mater. 11, 3227 (1999). 475. M. E. Hagerman, S. J. Salomone, R. W. Herbst, and A. L. Payeur, Chem. Mater. 15, 443 (2003).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Dynamic Processes in Magnetic Nanostructures B. Hillebrands, S. O. Demokritov Technische Universität Kaiserslautern, Kaiserslautern, Germany

CONTENTS 1. Introduction 2. Preparation of Patterned Magnetic Structures 3. Spin Wave Spectrum in Restricted Geometries 4. The Brillouin Light Scattering Technique 5. Spin Wave Quantization in Stripes 6. Static Magnetic Coupling in Arrays of Rectangular Dots 7. Dynamic Interdot Coupling in Arrays of Circular Dots 8. Spin Wave Wells 9. Conclusions Glossary References

1. INTRODUCTION The subject of this review is the dynamic properties of small magnetic elements made by patterning magnetic films and multilayers. Such elements generally receive very high attention, as they constitute, for example, magnetic sensors in magnetic read heads in hard discs or in magnetic random access memory. Because of the demand for higher speed of such sensors, the physical properties of fast magnetization processes came into the focus of interest in the last years. As one fundamental property, and this is the main subject of this review, the eigenexcitation spectrum of such elements in the linear excitation regime is of central interest. The availability of this basic knowledge is mandatory to understand dynamic processes, in particular those involved in fast switching of the magnetization direction, in general. Magnetic elements show a large variety of dynamic excitations. The most fundamental one is the uniform precession, where the magnetization precesses about a common direction given by the direction of the internal field. Such a ISBN: 1-58883-058-6/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

precession is sensed by the method of ferromagnetic resonance, a technique known since the 1970s. A uniform precession can be regarded as a spin wave with zero wavevector. Modes with nonzero wavevector may exist as well. However, boundary conditions at the edges of the elements exist, which will cause a characteristic quantization into a discrete set of excitation modes, similar to the fundamental mode and the overtone modes of a guitar string. It is the subject of this chapter to discuss the fundamental properties of the magnetic excitation spectrum in patterned magnetic elements. There exists a characteristic peculiarity in magnetic systems: since the elements are magnetized, they generate a magnetic stray field outside the elements. The interaction of the stray field with the internal distribution of magnetization results in an intrinsic inhomogeneity of the internal field distribution. Therefore magnetization dynamics demands an explicit consideration of spatially varying internal fields, which, on one hand, makes the problem more complex, on the other hand, results in new phenomena, like the localization of modes in certain regions within the element. Magnetic phenomena in small objects are driven by two interactions: the exchange interaction, which inside the element leads to ferromagnetic order, and dipolar interaction. Stray fields usually play a very important role, and, although the stray field energy is orders of magnitude lower than exchange energy, we observe a large variety of effects, like coupling and particular modes. Exchange interaction, taking place on small distances below several nanometers only, appears in the spectra through the existence of shortwavelengths modes. The outline of the chapter is as follows. After a short description of the preparation of patterned magnetic structures, we will describe in detail the spin wave spectrum in a restricted geometry. One central tool is the Brillouin light scattering technique, which will be discussed in Section 4. With increasing degree of complexity, we next discuss the excitation spectrum in various restricted geometries, beginning with the quantization of spin waves due to the lateral edges in magnetic stripes (Section 5) and in

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 2: Pages (695–716)

696

Dynamic Processes in Magnetic Nanostructures

dots (Sections 6–8). As an important side step, we discuss the static magnetic coupling in arrays of rectangular dots (Section 6) and the dynamic interdot coupling in arrays of circular dots (Section 7). An important physical phenomenon is the appearance of localization effects of spin waves in small elements due to the inhomogeneity of the internal field. This effect, which is reminiscent of localization effect in a potential in quantum mechanics, will be discussed in detail in Section 8. Section 9 concludes this review.

2. PREPARATION OF PATTERNED MAGNETIC STRUCTURES The technology for fabrication of high-quality patterned magnetic structures with lateral extensions on the micrometer, submicrometer, and nanometer length scale has been perfected to a remarkable degree in the past decade [1–3]. Lateral magnetic structures are conveniently fabricated from magnetic films using lithographic patterning procedures. In the following, for simplicity, we will call structures with one restricted lateral dimension “stripes,” and those with two restricted dimensions circular “dots” or rectangular “elements” following the usual conventions, although, as will be one of the subjects of this review, no real reduction of dimensionality is given since both in “stripes” and “dots” the magnetization is not constant over each magnetic object along a direction of restricted dimension. It is usually extremely difficult to study a single magnetic element, since it challenges the sensitivity of the measurement setup [4]. To avoid this problem the elements under investigation are usually assembled in arrays. Changing the distance between elements in an array, one can additionally investigate the interaction between the elements. Metallic Fe [5], FeNi [6–8], or Co [9, 10] films are mainly used for patterning. For fundamental studies FeNi films are preferable due to their smallness of the coercive field, the weak intrinsic magnetocrystalline anisotropy, and magnetostriction. The saturation fields of FeNi films are small, and the vanishing intrinsic anisotropy does not inhibit the observation of sometimes minute anisotropy effects caused by, for example, the shape of the elements or interactions between elements. However, patterned structures made on the basis of Co films allow one to investigate the interesting case of perpendicularly magnetized dots or stripes [9]. The patterning process is most often performed by means of electron beam lithography (EBL) [11–13], X-ray lithography (XRL) [14], and laser interference lithography (LIL) [15], followed by ion beam etching for pattern transfer. High-quality samples can be fabricated using all three processes, but each of them has its characteristic advantages and disadvantages. The EBL technique is very versatile, and can provide a very high resolution down to 10 nm, but due to its serial character it is time consuming, in particular for large patterned areas. Figure 1 shows scanning electron micrographs of a dense array of permalloy magnetic rectangular elements demonstrating the quality which can be achieved by EBL. In the XRL technique a resist is exposed to synchrotron radiation through a metallic mask. The technique possesses the potential to pattern large areas for standard magnetic

4ccV Spel Magn Det WD Exp 10.0 kV 8.0 21147× TLD 4.7 9428 sample:

2 µm

Figure 1. Scanning electron micrograph of an array of rectangular 1 × 175 m2 permalloy elements with a spacing of 0.1 m. The structure is obtained by EBL. Reprinted with permission from [64], S. O. Demokritov and B. Hillebrands, J. Magn. Magn. Mater. 200, 706 (1999). © 1999, Elsevier Science.

measurements, and one can achieve a high fabrication yield for systematic studies versus, for example, film thickness or material composition. A resolution as low as 50 nm can be achieved using XRL [16] which allows the exploration of mesoscopic magnetic phenomena. The key step in XRL is the mask fabrication, which is performed using EBL. This is a rather complicated and time-consuming procedure. However, once prepared, a mask can be used for XRL patterning of hundreds of samples. Thus, XRL is a fast and convenient method, but the need of a synchrotron radiation source and a mask fabrication step restricts its applicability. Figures 2 and 3 show scanning electron micrographs of an array of stripes and circular dots, respectively, prepared by XBL. LIL is a fast and easy process well suited for generating periodic arrays of stripes or ellipsoidally shaped dots on large areas with high coherency. Here the magnetic film coated with a photoresist is exposed to the interference pattern produced by two laser beams. Double exposure with a rotation of the sample of 90 or 120 generates patterns with fourfold or hexagonal symmetry. Very recently ion beam [3, 17] and electron beam [18] irradiation were used for patterning of the magnetic properties of Co/Pt multilayers. The advantage of both techniques is that the patterning process does not affect the meso- and macroscopical roughness of the surface and the optical properties of the film. The ion beam irradiation technique takes advantage of a low-density displacement of atoms by light ion irradiation. The microscopical mechanism which causes the changes of the magnetic properties of a film by electron beam bombardment is still not clear [18, 19]. Ion/electron-beam-induced magnetic patterning allows one to create adjoining regions with very different magnetic properties, such as perpendicular versus in-plane magnetization.

Dynamic Processes in Magnetic Nanostructures

697 The dipole-exchange spin wave spectrum in an unlimited ferromagnetic medium is given by the Herring–Kittel formula [22]

z

=

y

x

Acc.V Spot Magn Det WD 30.0 kV 1.0 10000x SE 14.8 CNRS / L2M

2 µm

Figure 2. Scanning electron micrographs of permalloy stripes with a width of 1.8 m and a separation of 0.7 m. The structure is obtained by XBL. The orientation of the Cartesian coordinate system used for calculations is shown as well. Reprinted with permission from [64], S. O. Demokritov and B. Hillebrands, J. Magn. Magn. Mater. 200, 706 (1999). © 1999, Elsevier Science.

3. SPIN WAVE SPECTRUM IN RESTRICTED GEOMETRIES The problem of the calculation of a spin wave spectrum for an axially magnetized infinite ferromagnetic stripe with a rectangular cross section has never been solved analytically (see, e.g., comments in [20]). However, in the particular case of a thin stripe with d ≪ w, where d is the thickness of the stripe and w is its width, the spectrum of long-wavelength magnetic excitations can be calculated approximately using the theory of dipole-exchange spin waves in a magnetic film [21].

 2



H+

2A 2 Q MS



H+

2A 2 Q + 4MS sin2 Q MS

1/2

(1)

where  is the gyromagnetic ratio, A is the exchange stiffness constant, H and MS are the applied magnetic field and the saturation magnetization both aligned along the z-axis, Q is the three-dimensional wavevector, and Q is the angle between the directions of the wavevector and the magnetization. In a magnetic film with a finite thickness d, the spin wave spectrum is modified due to the fact that the translational invariance of an infinite medium is broken in the vicinity of the film surfaces. An approximate expression for spin wave frequencies of a film can be written in the form, analogous to Eq. (1) (see Eq. (45) in [21]): =

 2



H+

2A 2 Q MS



H+

1/2 2A 2 Q +4MS · Fpp q d MS (2)

where 2

Q =

qy2

+

qz2

+



p d

2

=

q2

+



p d

2

(3)

Here the normal to the film surface points along the x-direction. q is the continuously varying in-plane wavevector, Fpp q  d is the matrix element of the magnetic dipole interaction, and p = 0 1 2    is a quantization number for the so-called perpendicular standing spin waves. Equation (3) is obtained under boundary conditions of “unpinned” spins on the film surfaces for the dynamic part m of the magnetization:    m    =0 (4)  x  x = ±d/2

For a general description of the spin wave modes one can use a more complicated boundary condition instead of the above one [23]: ±

1 µm

Figure 3. Scanning electron micrographs of the permalloy dots with a diameter of 1 m and a separation of 0.1 m. The structure is obtained by XBL. Reprinted with permission from [64], S. O. Demokritov and B. Hillebrands, J. Magn. Magn. Mater. 200, 706 (1999). © 1999, Elsevier Science.

m + Dm x = ±d/2 = 0 x

(5)

where the so-called pinning parameter D is determined by the effective surface anisotropy, kS , and the exchange stiffness constant A: D = kS /A. In the following the approximation D = 0 is justified by the small values of anisotropies in permalloy. If the spin wave is propagating in the film plane, but perpendicular to the bias magnetic field (qz = 0, q = qy ), the expression for the matrix element of the dipole-dipole interaction Fpp q d has the form [21] Fpp = 1 +

4MS Ppp 1 − Ppp  2A H+M Q2 S

(6)

Dynamic Processes in Magnetic Nanostructures

698 where the function Ppp q  d for the lowest thickness mode p = 0 is given by [21] P00 =

1 − exp−q d q d

(7)

The explicit expressions for Ppp⊥ when p > 0⊥ are also given in [21] (see Eq. (A12) in [21]). In a long-wavelength approximation (q d < 1) neglecting exchange (A = 0), the dispersion equation for the lowest thickness mode (p = 0) obtained from Eqs. (6), (7) gives results that are very similar to the results obtained by Damoun and Eshbach [24]: DE =

 · H · H + 4MS  + 2MS 2 · 1 − e−2q d  2

1/2

(8)

Thus, if the film is magnetized in-plane and q ⊥ MS , the spin wave modes described by Eqs. (3), (4), (6), (7) can be divided into the dipole dominated surface mode (p = 0, so-called Damon–Eshbach (DE) mode) and exchange dominated, thickness- or perpendicular standing spin wave (PSSW) modes (p > 0). The frequency of the former is determined by either Eq. (2) (with p = 0) or Eq. (8), while the frequencies of the PSSW modes (p > 0) can be approximately calculated from the expression p =

 2

  2A 2 2A p 2 q + MS MS d     4MS /H 2 2 2A +H q · H+ MS p/d   1/2 2A p 2 + 4MS + MS d



H+

(9)

which is obtained from Eq. (2) in the limit q d ≪ 1 using the expressions for the dipole-dipole matrix elements Fpp q d calculated in [21]. It is clear from Eq. (9) that the p q  dependence is rather weak for q ≪ p/d. In the general case of q d > 1, a numerical approach is usually used for the determination of the spin wave frequencies [25]. Equations (3), (6)–(9) provide a proper description of the spin wave frequencies apart from the intervals of mode crossing 0 ≈ p , where an essential mode hybridization takes place. In these regions numerical calculations are also necessary to obtain the correct spin wave spectrum of the film [21, 25]. Let us consider a magnetic stripe magnetized in-plane ˆ along the z-direction and having a finite width w along the ˆ y-direction as shown in Figure 2. A boundary condition similar to Eq. (4) at the lateral edges of the stripe should be imposed:    m    =0 (10)  y  y = ±w/2 An additional quantization of the y-component of q is then obtained: qy n =

n w

(11)

where n = 0 1 2    . Using Eqs. (2), (8), and (9) and the quantization expression (11), one can calculate the frequencies of these so-called width (or laterally quantized) modes. The profile of the dynamic part of the magnetization m in the nth mode can be written as    w w w mn y = an · cos qy n y + (12)  − 0 very well, and for the n = 0 mode a reasonable agreement is achieved. Since the group velocity Vg = 2/q of the DE spin wave (cf. Eq. (8)) decreases with increasing wavevector, the frequency splitting of neighboring, width-dependent discrete spin wave modes, which are equally spaced in q-space qy n = n/w, becomes smaller with increasing wavevector qy n , until the mode separation is smaller than the frequency resolution in the BLS experiment and/or the natural line width, and the splitting is no longer observable in Figures 7 and 8. Although the spatial distribution and the frequencies of the observed modes are well reproduced, the above approach ignores the correction of dynamic dipole fields due to the finite width of the stripes. Since the magnetic dipole interaction is a long-range one, the strength of the field at a given position is determined not only by the amplitude of the dynamic magnetization in the vicinity of this position, but also by the spatial distribution of the dynamic magnetization far from this position. Therefore, due to finite size effects caused by the side walls, the dipole fields accompanying a quantized mode in the stripe differ from those of the DE mode of the infinite film with the same wavevector. This correction, which is small in any case for the stripes with a high aspect ratio, is negligible for all modes with sufficiently high quantum numbers (i.e., for modes with either p or n larger than zero). It is, however, observable for the lowest, uniform mode (p = 0, n = 0). The finite size effect can be easily taken into account for long stripes with ellipsoidal cross section since the dynamic dipole field is homogeneous in this case. The corresponding frequency is given by the Kittel formula [38]: =

 · H + Nx − Nz  · 4MS  2 · H + Ny − Nz  · 4MS 

1/2

(17)

where Nx , Ny , and Nz are the demagnetizing factors along ˆ and z-direction ˆ y-, ˆ the x-, in the stripe cross section (note here that the demagnetizing factor along the stripe, Nz , is negligible). In the particular case of a stripe with a rectangular cross section Eq. (17) is not exactly applicable, since the dynamic demagnetizing field is not homogeneous. However, since the thickness of the stripe is much smaller than its width d ≪ w, one can consider this field to be approximately homogeneous over the main part of the stripe in agreement with the analytical solution for demagnetizing fields of a rectangular prism found by Joseph and Schlömann [39]. The corresponding demagnetizing factors are: Ny = 2d/w, Nx = 1 − Ny , Nz = 0. The calculated value of Ny is in good agreement with that obtained from static measurements. The frequency of the lowest mode calculated on the basis of Eq. (17) using the above demagnetizing factors is shown by a dotted horizontal line in Figures 7 and 8. The agreement with the experiment is convincing. It is much more complicated to take into account the finite size effects for the nonuniform modes. However, from the

qualitative considerations it is obvious that the correction of the demagnetizing fields (or demagnetizing factors) due to the finite size of the stripe rapidly decreases with increasing mode number. Therefore, good reproduction of the mode frequencies for n > 0 just on the basis of Eq. (17) is not surprising.

6. STATIC MAGNETIC COUPLING IN ARRAYS OF RECTANGULAR DOTS Before we discuss in more detail the dynamics in films patterned in two dimensions, we first need to visit the static properties. This is the subject of this section. We present a study of the static magnetic properties of dense arrays, consisting of micron-size rectangular magnetic elements with small interelement spacings. Special attention is paid to the influence of the magnetic dipole coupling on the magnetization curves and magnetic domain configurations. The rectangular elements with a thickness of 35 nm and lateral dimensions of 1 × 1 m2 , 1 × 12 m2 , and 1 × 175 m2 respectively were prepared for these studies. The spacing d between the elements was always chosen the same in both lateral directions. In addition several arrays with the same element size 1 × 175 m2 ) and different interelement spacings () = 01 m, 0.2 m, 0.5 m, and 1 m) were prepared. The elements were arranged in arrays with overall dimensions of 800 × 800 m2 . The high quality of the patterning process is evident from Figure 1, where a scanning electron micrograph of the elements with lateral dimensions of 1 × 175 m2 and a spacing of 0.1 m is shown. Measurements of the magnetization curves were performed using longitudinal magneto-optic Kerr effect (MOKE) magnetometry with the magnetic field applied in the plane of the sample. The measured signal in this case is proportional to the in-plane magnetization component parallel to the applied field. The domain structures of the elements were monitored at room temperature by magnetic force microscopy (MFM). For imaging, two MFM microscopes, Solver MFM from NT-MDT and NanoScope IIIa from Digital Instruments, were used. The MFM tips were magnetized along the tip axis, that is, normal to the sample surface, making the MFM sensitive to the vertical component of the stray fields. The unpatterned films are characterized by square shape magnetization loops with coercivities less than 2 Oe, which reflect the high quality of the films. Measured at different in-plane orientations of the applied magnetic field, the magnetization curves demonstrate quasi-isotropic properties of the films, an in-plane anisotropy field being below 1 Oe. The magnetization curves of the patterned arrays show higher coercivities and much higher saturation fields between 30 and 250 Oe. Typical examples of the magnetization curves measured by MOKE for the magnetic field applied along the long and short sides of the rectangles are presented in Figure 10. The pronounced difference between the two curves reflects the anisotropic nature of static magnetic properties of the array. An arrow in Figure 10 indicates the experimentally determined saturation field HS as an averaged value over two saturation fields HS1 and HS2 measured for different parts of the hysteresis loop. Such an averaging compensates the effect of the hysteresis.

Dynamic Processes in Magnetic Nanostructures

704

Hs Hs2

h.a.

Kerr signal (a.u.)

Hs1

e.a.

-150

-100

-50

0

50

100

150

H (Oe)

Figure 10. Typical magnetization curves of the array of 1 × 175 m2 elements with a spacing of 0.1 m measured along their easy (e.a.) and hard (h.a.) axes using MOKE. The arrow indicates the value of the measured saturation field for the hard-axis curve determined as an averaged value of HS1 and HS2 measured for different parts of the hysteresis loop (shown is only HS for the hard-axis curve).

Figure 11 shows the obtained saturation field versus the in-plane angle of the applied field for the arrays with different lateral dimensions of the elements, but with a fixed separation between the elements of 0.1 m. For the arrays of rectangles a twofold anisotropy with the easy axis lying along the long axis of the elements is clearly seen. The experimentally determined magnitude of the saturation field depends on the lateral dimensions of the elements, on the in-plane orientation of the applied field with respect to the rectangle sides, and on the distance between the elements in the array. To saturate the sample the applied field needs to be higher than the demagnetizing field which is determined by the

shape of the elements and the interelement distance. The effect is attributed to the element shape, as corroborated by a larger anisotropy observed for the elements with a higher lateral aspect ratio. For the array of squares a weak fourfold anisotropy caused by the inhomogeneity of the demagnetizing field is observed (configurational anisotropy), while the continuous film is almost isotropic in the plane, as is seen from the solid circles in Figure 11. Although the observed magnetic anisotropy is strongly correlated with the shape of a single element, it depends on the interelement distance as well. The circles and the squares in Figure 12 show the saturation fields along the easy and hard axis (the long and short sides of the rectangles) measured as a function of the interelement spacing for the array of elements with the lateral size of 1 × 175 m2 . For “zero spacing,” that is, a continuous film, the data are shown for reference. A sizeable increase in the saturation field with increasing spacing and nearly a constant saturation field at higher spacing values demonstrate the fact that the interaction between the elements influences drastically the magnetization process of the array. For large spacings, that is, well above 1 m, the elements can be considered as noninteracting. In this case the saturation field of the arrays is determined by the demagnetizing field of an individual element. With decreasing separation between the elements, the demagnetizing fields of the neighboring rectangles partly compensate each other and the saturation field of the array is decreasing. The obvious limit of zero spacing is a nonstructured film with a very low saturation field, which is determined by the intrinsic coercivity of the film. The influence of the magnetostatic interaction between the elements can be understood from the analysis of the magnetic dipole field of the interacting elements. For a uniformly magnetized array of rectangular elements this can be done rigorously using the approach developed in [40]. For calculating the dipolar fields we assume Cartesian coordinates with the x- and y-axes lying along the rectangle sides 300 easy axis hard axis

100 [01] H

Φ

200

Hs (Oe)

Hs (Oe)

80

60 100 40

0 0,0

20

0,2

0,4

0,6

0,8

1,0

spacing (µm) 0

45

90

135

180

225

270

315

360

in-plane rotation angle Φ (deg)

Figure 11. The saturation field HS as a function of the in-plane angle measured with respect to the short axis of the element: —unstructured film; —array of square 1 × 1 m2 elements with spacing ) = 01 m; —array of 1 × 175 m2 elements with spacing ) = 01 m; and —array of 1 × 12 m2 elements with spacing ) = 01 m.

Figure 12. Measured (symbols) and calculated (lines) values of the saturation field HS for the array of 1 × 175 m2 rectangular elements as a function of the interelement spacing ).  and the solid line—the field is along the short edge of the element (hard axis),  and the dashed line—the field is along the long edge of the element (easy axis). Reprinted with permission from [64], S. O. Demokritov and B. Hillebrands, J. Magn. Magn. Mater. 200, 706 (1999). © 1999, Elsevier Science.

Dynamic Processes in Magnetic Nanostructures

705

d6x y z′  = dz′





dkx

−

× e−



 −



dky 

kx2 +ky2 z−z′

iMky kx2 + ky2

˜ x  ky e−ikx x e−iky y 7k

(18)

where the Fourier component of the two-dimensional array structure 7x y is  y 1   ikxx e dx eiky dy7x y 2 − −    nL +l /2  x x  1 x eikx dx = 2 n = − nLx −lx /2     mL +l /2 y y  y iky dy e ·

7kx  ky  =

(19)

m = − mLy −ly /2

with the translation vectors of the array in the x- and ydirection Lx = lx + ) and Ly = ly + ). Substituting 7x y in [1] and integrating over kx and ky , one obtains d6x y z − z′  =

 ie−iqx x e−iqy y 8M dz′  Lx Ly n m = − qqx     qy ly −q z−z′ q l × sin x x sin e 2 2

(20)

 n, qy = 2 m and qx2 + qy2 are the vectors of where qx = 2 Lx Ly the reciprocal lattice of the array. In order to calculate the potential of the entire structure, the expression (20) should be integrated over z′ from −h/2 to h/2. For −h/2 < z < h/2, the result of the integration is written as follows: 6xyz =

    qy ly 8Mlx ly  ie−iqx x e−iqy y qx lx sin Lx Ly n m = − q 2 qx 2 2 (21)

×2−e−qz+h/2 −eqz−h/2

The magnetic field can be found by differentiating this expression in accordance with H = −8 9. The y-component of the field Hy x y z, which is antiparallel to the magnetization and is therefore decisive for the remagnetization process, can be written as follows:    qy ly 6 lx 8M  Hy x y z = − = sinqy y cos y Lx L y m = l qy 2

× 2 − eqy z+h/2 − eqy z−h/2 −

 

n m = l

4qy q 2 qx

    qy ly q l × sinqx x sinqy y cos x x cos 2 2

(22) × 2 − eqy z+h/2 − eqy z−h/2 The calculated profile of the demagnetizing field over a single element in the array of the 1 × 175 m2 elements with ) = 01 m is shown in Figure 13 for the magnetization lying along the short side of the element. Since the elements have a rectangular, nonellipsoidal shape, the demagnetizing field is not homogeneous. It is significantly enhanced in the zones near the edges of the element perpendicular to the magnetization. Thus, the saturation of the entire element occurs at higher fields compared to the demagnetizing field in the center of the element Hd0 . The process of the element magnetization can be divided into three steps: (i) the applied field is below Hd0 —a domain walls movement and/or a rotation of the magnetization in the domains take place; (ii) the applied field exceeds Hd0 —the main part of the element is magnetized into the new field direction; (iii) the applied field approaches and then exceeds the much higher value of the demagnetizing field near the edges—the entire element is saturated. Thus, from the experimental point of view, a fast increase of the element magnetization during step (ii), followed by a very slow saturation at high fields, can be considered as a partial saturation of the main part of the element. The calculated demagnetizing field Hd0 in the center of the element, that is, for x = y = 0, averaged over z, is plotted in Figure 12 as a function of the distance between the elements ). This field mainly determines the experimentally measured saturation field, which is also shown in Figures 12 and 13. As it is seen in Figure 12, both fields gradually increase with increasing ). This increase reflects the fact that magnetic poles of opposite sign, which belong to neighboring rectangles, become more and more separated, reducing 2,0

1,5

Hdip (kOe)

of the size lx and ly , respectively, and the z-axis lying along the surface normal, as is shown in Figure 1. The origin of the system is placed in the center of an element, that is, the element occupies the volume of −lx /2 < x < lx /2, −ly /2 < y < ly /2, and −h/2 < z < h/2. The magnetization is assumed to be homogeneous across the element and oriented along the y-direction. The scalar potential which is produced by an infinitely thin layer of the array, lying within the xy-plane at height z, can be found using the following expression:

1,0

0,5

0,0 -0,50

-0,25

0,00

0,25

0,50

y (µm)

Figure 13. The profile of the magnetic dipolar field across the element at x = 0 for the array of 1 × 175 m2 elements with spacing ) = 01 m for the magnetization lying along the short side of the element. The dipolar field is averaged over the thickness of the element h. Note that the field is oriented opposite to the magnetization direction. The dashed horizontal line indicates the experimentally measured value of the saturation field.

Dynamic Processes in Magnetic Nanostructures

706 the diminishing effect on the resulting dipolar field. In the limit of large separations the calculated dipolar field of the array saturates and it becomes characteristic for an individual rectangular prism [41]. The calculated dipolar field Hd0 reproduces qualitatively the behavior of the measured saturation field: both fields grow with increasing ) and become almost constant for ) > 07 m. The lack of the quantitative agreement is a consequence of the inhomogeneity of the demagnetizing field over the element dimensions, as it is discussed above. To take this into account one needs to solve the problem self-consistently implying a nonuniform distribution of the magnetization. For a deeper insight into the magnetic ground state of a single rectangular element, MFM images of the samples were measured at zero field. It was found that the remanent domain structure depends drastically on the demagnetization history. If a moderate magnetic field of 100–150 Oe is applied along the long edge of the element, the so-called S- and Cdomain states are observed. If the field is applied along the short edge of the element, the Landau state is observed on the majority of the elements. In addition the MFM imaging process slowly changes the domain pattern. After several imaging cycles many elements show a diamond-like domain pattern. This pattern completely avoids 180 walls and consists of 90 walls only. The observed domain patterns are presented in the upper panel of Figure 14. It is worth noting here that the MFM tips covered by carbon caps do not affect the domain patterns. For a relatively large separation between the elements the observed domain patterns can be simulated by considering the micromagnetic behavior of a single element. The numerical approach, based on a direct solution of the Landau–Lifshitz–Gilbert torque equation, was applied for this purpose (OOMMF program package [42]). It is evident from Figure 14 that the modeling demonstrates excellent agreement between the calculated and measured domain configurations (compare the top and bottom

panels in Fig. 14). As was found by the modeling, the three domain configurations have different energies, the lowest energy given by the diamond-like configuration. This implies that the S-domain structure and the Landau structure are metastable and can be brought to the diamond-like state by a corresponding perturbation. In our case the stray field of the MFM tip is such a perturbation. A significant interelement interaction in the arrays with small spacings becomes obvious from the MFM images. Figure 15 shows two MFM images (65 × 9 m2 ) for the array of 1 × 175 m2 elements with ) = 01 m at two different external fields. In order to magnetize the array, a high field of 120 Oe was applied along the long side of the elements before imaging. As is evident from Figure 15, the magnetization close to the short edges of the elements tends to rotate towards the direction parallel to these edges, forming the S-domain structure. At high fields (not shown) these edge domains are rather narrow, but in remanence they cover extended triangular areas, as can be seen in Figure 15. The resulting magnetic charges are positioned close to the short edges of the elements and are imaged in the dark and light shades. Their respective positions for each individual element indicate the dominant direction of the magnetization within the elements. Figure 15a shows the domain configuration, which was recorded after decreasing the external field and reaching a weak reverse field

a

H = -48 Oe b

H = -64 Oe E  3.7 pJ

E  1.3 pJ

E  1.2 pJ

Figure 14. Comparison between the experimental (upper panel) and simulated domain patterns (lower panel) of 1 × 175 m2 rectangular elements. The total energies of the pattern obtained in the simulations are also shown.

Figure 15. MFM images (65 × 9 m2 ) of the array of the 1 × 175 m2 rectangular elements recorded during a magnetization reversal process. The sample was magnetized at 120 Oe. (a) MFM image taken at −48 Oe. The gray rectangle indicates a single element; the white arrows indicate the main orientation of the magnetization within the element. (b) MFM image taken at −64 Oe.

Dynamic Processes in Magnetic Nanostructures

of 48 Oe. The fact that the magnetization reversal process has been already initiated manifests itself in the second row of elements from the bottom. Here the entire row has been switched to the opposite magnetization direction. An increase in the reverse field up to 64 Oe results in switching all the rows excluding the second row from the top, which remains in the initial configuration. Figure 15 demonstrates a significant intrarow interaction, which keeps the elements within one row in a common magnetization direction. In this case the inter-row interaction strength is much weaker. Nevertheless, it can be observed after ac demagnetization of the sample. Figure 16a and b displays two zero-field MFM images, recorded after demagnetizing the sample in an ac field along the long and short axes of the elements, respectively. As is seen from Figure 16a, each individual element still carries a net magnetization and shows the triangular edge domains, similar to those in Figure 15. A significant intrarow interaction is evident from the fact that all the elements in a row are found in the same magnetization direction, that is, each row still carries macroscopic nonzero magnetization. By comparing the black-white contrast for the rows one finds, however, that the magnetization alternates in most cases from pointing to the right to pointing to the left for the neighboring rows. As a result, although the elements and even the rows still carry the net magnetization, the overall magnetization of the sample is compensated to zero by the alternating magnetization directions of the rows. The presence of an inter-row interaction is also evidenced by a high number of cases where the edge domains of four neighboring islands

a

b

Figure 16. Remanence MFM images (65 × 9 m2 ) of the array of the 1 × 175 m2 rectangular elements after demagnetizing in ac fields. (a) Demagnetizing field is along the long edge of the element. (b) Demagnetizing field is along the short edge of the element.

707 show a strong correlation between the rows. They tend to form a structure with an alternating black-white contrast, which allows an effective flux closure on a local scale. One might expect that a configuration with a demagnetized state for each single element is energetically more favorable. As is obvious from Figure 16b, such a situation can indeed be generated after ac-demagnetizing the sample along the short axis. Now one finds the elements in the Landau state with some of the elements displaying a single cross-tie to be present within the central 180 wall, as it was also seen for isolated permalloy elements [43]. Since the net magnetization of the Landau state is smaller than that of the S-state, the interelement coupling is not apparent in Figure 16b.

7. DYNAMIC INTERDOT COUPLING IN ARRAYS OF CIRCULAR DOTS Up to now there exist very few studies of spin waves in magnetic dot arrays [5–7, 44]. This is definitely connected to the fact that, on one hand, it is much more difficult to prepare dots with a well-defined shape. On the other hand, the magnetic signal from dots is lower than that from stripes simply due to the lower coverage of the surface by the magnetic material. Grimsditch et al. [5] investigated submicron Fe magnetic dot arrays by BLS with dots of ellipsoidal shape. They showed that the shape anisotropy of individual dots is a dominant source of anisotropy, measured both by static magnetometry and BLS. The measured spin wave frequencies are in good agreement with values calculated on the basis of isolated ellipsoids using Eq. (17). No interdot coupling was observed. Hillebrands et al. [6, 44] investigated spin wave properties of square lattices of micron-sized circular dots of permalloy with varying dot separations. Different samples comprising circular dots arranged in a 1 × 1 mm2 square lattice with a diameter/separation of 1/0.1, 1/1, 2/0.2, and 2/2 m, respectively, patterned into 500- and 1000-Å-thick films were prepared. Special care was taken to avoid a touching of neighboring dots even for the smallest separation. This was confirmed by depth profile measurements. The measured spin wave frequencies of the different dot arrays as a function of the applied field are shown in Figure 17. For each dot array the spin wave frequencies decrease with decreasing field, and they disappear below certain critical applied fields. The strong reduction of the spin wave frequencies with decreasing aspect ratio L/d is also seen in Figure 17. This is caused by size-dependent demagnetizing fields, that is, the demagnetizing factor N of each magnetic dot decreases with the aspect ratio. The solid lines in Figure 17 are calculated on the basis of Eq. (8), where the applied magnetic field is substituted by the internal field in the dot. The latter is calculated using demagnetizing factors of spheroids with axial ratios taken from the aspect ratios of the magnetic dots. Although spheroids are a crude approximation to the real three-dimensional shape of the dots, the calculation is in reasonably good agreement with the experimental data for H > N · 4Ms . To investigate the problem of an in-plane interdot coupling, the spin wave frequencies were measured as a function

Dynamic Processes in Magnetic Nanostructures

708

anisotropy constant, the free energy expression is

22 continuous film

Spin Wave Frequency (GHz)

21

F = K 4 sin2 6 · cos2 6

d/L=0

2/0.2 nm 2/2 nm

20

1/0.1 nm

d/L=0.05

1/1 nm

19 18

d/L=0.1

17 16 L

15

d

14

0,0

0,5

1,0

1,5

2,0

2,5

Magnetic Field H (kOe)

Figure 17. Measured spin wave frequencies of the dot arrays with a dot thickness of d = 100 nm as a function of the magnetic field. The solid lines represent the calculated frequencies on the basis of Eq. (8) with substitution of the applied field by the internal field in the dot. Reprinted with permission from [64], S. O. Demokritov and B. Hillebrands, J. Magn. Magn. Mater. 200, 706 (1999). © 1999, Elsevier Science.

Spin Wave Frequency (GHz)

of the angle of the in-plane applied field, 6H , with respect to a reference [10]-direction of the lattice arrangement. For the smallest dot separations of 0.1 m a fourfold anisotropic behavior was found, which was seen neither for larger dot separations nor for the unpatterned reference film. This is displayed in Figure 18 for the 1/0.1 m and, for comparison, for the 1/1 m lattices of the sample of 100 nm thickness at an applied field of 1 kOe. Note here that the easy axes (maximum frequencies) of the observed anisotropy are along the [11]-directions of the lattice. To determine quantitatively the

17

15

16

14

15

13

0

45

90

135

180

In-plane Angle ΦH (deg)

Figure 18. Dependence of the spin wave frequencies on the in-plane direction of the applied field for the 1/0.1 mm (full squares) and, for comparison, for the 1/1 mm (open squares) dot arrays of 100 nm thickness. The solid line is a fit for the data. Note that maximum values of the spin wave frequencies indicate an easy axis. Reprinted with permission from [64], S. O. Demokritov and B. Hillebrands, J. Magn. Magn. Mater. 200, 706 (1999). © 1999, Elsevier Science.

(23)

with 6 the angle between the direction of magnetization with respect to the [10]-direction, and K 4 the constant of a fourfold in-plane anisotropy. A model fit using Eq. (23) and a numerical procedure to calculate the spin wave frequencies [45] is displayed in Figure 18 as a solid line for the 1/0.1 m lattice. For both 1/0.1 m samples with thicknesses of 50 nm and 100 nm, the anisotropy contribution K 4 was determined for several values of the applied magnetic field. The obtained values of K 4 decrease with increasing field, and they saturate within the investigated field range at K 4 = −06 · 105 erg/cm3 , which corresponds to an effective anisotropy field Hani = 150 Oe, at the same reduced field value of about H /Hd = 5 with Hd the demagnetizing field. As a function of the reduced field, H /Hd , the data of both dot thicknesses fall onto one common curve within the error margins indicating that the coupling strength scales with the demagnetizing field. The observed fourfold anisotropy can be understood as being caused by a magnetic dipolar interaction between residual unsaturated parts of the dots. This effect is known as “configurational anisotropy” [46]. Because of the large distance of 0.1 m between the dots, a direct exchange mechanism via conduction electrons or via electron tunneling can be excluded. A dipolar interaction of completely magnetized dots also cannot account for the observed anisotropy, because the corresponding dipolar energy can be expressed as a bilinear form of the components of the magnetization vector. Such an expression can only yield an uniaxial, but not a fourfold anisotropy contribution, since in a bilinear form the direction cosines appear quadratic in highest order and add to a constant if a fourfold symmetry is given. But, if the dots are not completely saturated, the magnetic dipolar interaction energy cannot be expressed in the above bilinear form, and the fourfold anisotropy is no longer forbidden. The large observed decrease of the coupling anisotropy constant with increasing field, forcing the alignment of the magnetization with the field, corroborates this assumption. Similar effect of higher-order magnetic anisotropies caused by the shape of the dot has been recently observed in static measurements [46]. Cherif et al. [7] studied magnetic properties in periodic arrays of square permalloy dots. The spin wave frequencies were found to be sensitive to the size of the dots and, for the smallest structures, to the in-plane direction of the applied field. While the size dependence can be reasonably explained to originate from the demagnetizing field effect, the in-plane anisotropy should be discussed in more detail. The authors observed a fourfold anisotropy in the spin wave frequencies as a function of the in-plane angle of the external field. The spin wave frequencies show maxima for the magnetic field applied along the edges of the squares, indicating easy axes of anisotropy along the edges. Contrary to the results of Hillebrands et al. [6, 44], the authors did not demonstrate that the observed anisotropy is caused by interdot interaction; instead they argue that it may be a single-dot effect. In fact, dots of square shape possess a lower symmetry compared to circular shaped dots. Therefore, shape

Dynamic Processes in Magnetic Nanostructures

709

intensity (a.u.)

2000

1500

1000

region of interest x8

500

0

0

5

10

15

20

frequency shift (GHz)

Figure 19. Obtained BLS spectrum for a square array of dots with a dot diameter of 2 m and a dot separation of 0.2 m for the transferred wavevector q = 021 · 105 cm−1 with an external field of 600 Oe applied in-plane demonstrating the existence of several discrete spin wave modes. The peaks corresponding to laterally quantized modes are indicated by arrows. In the so-called region of interest the scanning speed was reduced by a factor of 8, increasing the number of of recorded photons by this factor. Reprinted with permission from [64], S. O. Demokritov and B. Hillebrands, J. Magn. Magn. Mater. 200, 706 (1999). © 1999, Elsevier Science.

16

spin wave frequency (GHz)

effects (corner effects for the magnetization, confinement of the spin wave mode, etc.) can cause this anisotropy. The spin wave mode splitting caused by confinement of the spin waves and corresponding quantization of the wavevector was not observed as well. Jorzick et al. [47] investigated square lattices of circular permalloy micron dots, similar to those used in [6, 44]. The only difference was the smaller thickness of the dots of d = 40 nm. Figure 19 shows the anti-Stokes side of a typical BLS spectrum for q = 042 · 105 cm−1 of an array with a dot diameter of 2 m and a dot spacing of 0.2 m at H = 600 Oe. Similar to the results obtained from the studies of magnetic stripes, several peaks, corresponding to in-plane quantized spin wave modes, as well as an exchange dominated, perpendicular standing spin wave mode are seen in Figure 19. By changing the angle of light incidence the value of q is varied and the dispersion of the observed mode is obtained as demonstrated in Figure 20 for the array of dots with 2 m dot diameter and 0.2 m dot separation. As it is seen in the figure, at least seven discrete dispersionless modes are detected in the wavevector interval, q = 0–08 · 105 cm−1 . For large values of q (>1 · 105 cm−1 ) only two well-defined spin wave modes are observed. Using a numerical procedure [45], the frequencies of the DE mode and the first perpendicular standing spin wave mode for a continuous film of same thickness (d = 40 nm) were calculated. The results of the calculation are plotted in Figure 20 as full lines. A good agreement between the experimentally measured and the calculated dispersion curves for q > 1 · 105 cm−1 demonstrates the fact that for large transferred wavevectors 2/q ≪ D the measured dispersion converges to the dispersion of a continuous film of the same thickness. In the transition region between well-resolved multiple modes and the dispersion of the continuous film q  08–1 · 105 cm−1  the mode separation is close to the BLS

14

12

10

8

6 0,0

0,5

1,0

1,5

2,0

2,5

q|| (105 cm-1)

Figure 20. Obtained spin wave dispersion curve for a square array of dots with a dot thickness of 40 nm, a dot diameter of 2 m, and a dot separation of 0.2 m. An external field of 600 Oe was applied in-plane along the axis of the array lattice. The solid lines show the results of a spin wave calculation for a continuous film of the same thickness as the dots. The dashed line marks the calculated frequency of the uniform mode of a single dot. Reprinted with permission from [64], S. O. Demokritov and B. Hillebrands, J. Magn. Magn. Mater. 200, 706 (1999). © 1999, Elsevier Science.

instrumental line width, and therefore separate modes cannot be resolved. Broad peaks are observed in the BLS spectra. A decrease of the frequency splitting between the modes observed at high values of wavevectors has been found for magnetic stripes [35]. It is due to the fact that the group velocity Vg = )"/)q of the DE mode decreases with increasing wavevector. In the case of stripes, where the spin wave modes are characterized by quantized equidistant wavenumbers q n (see previous section), this fact necessarily leads to a monotonical decreasing frequency splitting between the neighboring modes. Two-dimensional quantization conditions in a circular dot are not so simple, as are those for the stripes. Therefore, the splitting between the quantized modes in the dots does not follow the simple rule, but still we observe a small splitting at high wavevectors in dots as well. The frequency of the uniform mode of a single disc (" = 735 GHz) calculated by means Eq. (17), using the values of the demagnetizing factors obtained from separate static magneto-optic Kerr magnetometry, is shown as a horizontal dashed line in Figure 20. The calculations clearly demonstrate that the two lowest quantized modes have frequencies near the frequency of the uniform mode. The five lowest spin wave modes of arrays of dots with the same dot thickness d = 40 nm as presented in Figure 20, but with a smaller dot diameter (D = 1 m), are shown in Figure 21 for two different dot spacings. From a comparison of the data presented in Figures 20 and 21 it is seen that the wavevector interval, where each mode is observed, scales approximately as the inverse dot diameter, D−1 . For example, the two lowest modes of the dots with the diameter of 2 m are observed in the wavevector interval q  0–05 · 105 cm−1 , whereas the same modes of the dots with the diameter of 1 m are observed at q  0–1 · 105 cm−1 . This result is in agreement with the theoretical analysis performed in [35], which has shown that the

Dynamic Processes in Magnetic Nanostructures

710

coupling restricts the density of dot arrays in, for example, magnetic memory applications. Figure 21 shows the frequencies of the spin wave modes, obtained on arrays having the same dot thicknesses and dot diameters, but with different interdot distances, ). A clear frequency upshift of the two lowest modes of the sample with ) = 01 m documents the existing interdot coupling.

14

12 11 10 9

8. SPIN WAVE WELLS

8 7 6 5 0,0

H = 600 Oe

0,5

1,0

1,5

2,0

q|| (105 cm-1)

Figure 21. Dispersion of the five lowest spin wave modes measured on two square arrays of dots with the same dot thickness of 40 nm and the same dot diameter of 1 m. Open symbols indicate data for a dot separation of 0.1 m, full symbols of 1 m. Note the difference between the frequencies of the lowest modes for the two samples. The dashed horizontal line indicates the calculated frequency of the uniform mode of a single dot. Reprinted with permission from [64], S. O. Demokritov and B. Hillebrands, J. Magn. Magn. Mater. 200, 706 (1999). © 1999, Elsevier Science.

light scattering intensity from a given spin wave mode confined to an island is determined by the Fourier transform of the mode profile over the island. Changing the thickness of the dots for a constant dot diameter (the data are not shown), one observes that the frequency splitting between neighboring modes decreases with decreasing dot thickness, in accordance with the fact that the group velocity of the Damon–Eshbach mode at a given q decreases with decreasing film thickness. The above presented experimental findings lead us to the conclusion that the observed dispersionless, resonancelike modes are the Damon–Eshbach spin waves, quantized due to lateral confinement in a single dot. Earlier studies of the magnetostatic modes of in-plane magnetized macroscopic discs using a ferromagnetic resonance technique [48] do not present any comparison of the experimental findings with the theory. Surprisingly, despite numerous publications on inhomogeneous magnetostatic modes in finite-size magnetic samples (see, e.g., [49] and references therein), there is no appropriate theoretical description of such modes up to now. This can be probably explained by the low symmetry of the problem. In fact, in his pioneering work, Walker [50] considered an axially magnetized spheroid. Due to axial symmetry, the analytic solution of the Walker equation is possible in this case and it can be expressed in terms of Legendre functions. The magnetization direction of an in-plane magnetized dot breaks the axial symmetry of the Walker equation, and, correspondingly, that of possible mode profiles [51]. The solution (analytical or numerical) of the Walker equation corresponding to an in-plane magnetized disc is still needed. Since the considered spin waves are the magnetostatic modes, and since the magnetic dipole interaction is longrange, the question of possible interdot mode coupling is of importance. Also, from the practical point of view, interdot

In this section we discuss dynamic excitations in small magnetic items (long stripes and rectangular elements) with a large inhomogeneity of the internal field. It is known that in a nonellipsoidal element the internal field decreases near the edges of the element [39]. In some cases zones with zero internal fields can be created [52]. The problem of inhomogeneous internal field has been avoided in the previously discussed geometry used for the studies of the spin wave quantization in stripes, since the external field was applied along the long axis of the stripes. In the following discussion we assume a Cartesian coordinate system in which the x-axis is perpendicular to the plane of the elements, and the y-axis is along the long axes of the stripes. The wave vector q is chosen along the z-axis, that is, perpendicular to the stripes. Figure 22 shows two typical BLS spectra obtained from an array of stripes of width

a)

BLS Intensity (a.u)

spin wave frequency (GHz)

13

b)

c)

-20

-15

-10

-5

Frequency (GHz)

Figure 22. BLS spectra obtained on the stripe array for q = 1 · 105 cm−1 at He = 500 Oe. Panels (a) and (c) correspond to different experimental geometries as described in the text. Panel (b) represents the spectrum of a unpatterned film.

Dynamic Processes in Magnetic Nanostructures

711

w = 1 m, a length of 90 m, and a distance between the stripes of 0.5 m for the external in-plane magnetic field He = 500 Oe for different orientations of the field. Spectrum (a) is obtained for He oriented along the y-axis, thus presenting the DE geometry, discussed above. Spectrum (b) represents the unpatterned film. Spectrum (c) is recorded with both q and He aligned along the z-axis. As it is seen in Figure 22, spectra contain several distinct peaks corresponding to spin wave modes. The high-frequency peaks indicated as PSSW can be easily identified as exchange dominated perpendicular standing spin wave modes also observed in unpatterned films. Their frequencies are determined by the exchange interaction and the internal field. To investigate the nature of the other observed excitations, the dispersion was measured by varying q. It is displayed in Figures 23 and 24 for both orientations of He . Figure 23 representing the DE geometry clearly demonstrates the lateral quantization of the DE spin waves, resembling a typical “staircase” dispersion, discussed in Section 5. The interval of the observation of each mode in the q-space

237 Oe, there are no spin waves with the frequency 1 = 45 GHz. Therefore, the lowest mode can only exist in the spatial regions in the magnetic stripe where 0 Oe < H < 237 Oe. The corresponding turning points are indicated in Figure 26 by the vertical dash lines. Thus, the lowest mode is localized in the narrow region 1 having their frequencies above 5.3 GHz are not strongly localized and

q min

q max

0 3

0

6

9

q(105cm-1)

Figure 25. Dispersion of plane spin waves in the BWVMS geometry at constant internal fields as indicated.

exist anywhere in the stripe where the internal field is positive (0 < z/w < 039). In the experiment they show a band, since the frequency difference between the r and r+1 modes is below the frequency resolution of the BLS technique. The above presented one-dimensional analytical approach is applicable to long magnetic stripes. To describe the spin wave modes of two-dimensional rectangular magnetic elements, micromagnetic simulations of nonuniform magnetic excitations in such elements have been performed [55]. The method used is based on the Langevin dynamics, and the time evolution of the magnetization distribution in the magnetic element, which is discretized into Nx × Nz = 100 × 180 cells with magnetic moments i , is simulated using the stochastic Landau–Lifshitz–Gilbert (LLG) equation [55]. The effective field Hieff acting on the ith moment consists of a deterministic part Hidet (which includes external magnetic field and the fields created by the exchange and dipolar interactions between different cell moments) and a fluctuating part Hifl t. The correlation properties of this fluctuating field, which is intended to simulate the influence of thermal fluctuations, may be quite complicated in

300 ∆Z

H (kOe)

H x y z = He − Nzz x y z · 4MS

H = 237 Oe 8

Frequency (GHz)

(laterally quantized modes) and for p = 1 and m = 0 (PSSW) are in good quantitative agreement with the results of the experiments as shown in Figure 23. The material parameters used are: 4MS = 102 kOe, A = 10−6 erg/cm2 , /2 = 295 GHz/kOe. If He ez , the effect of demagnetization due to the finite stripe width is very large, and the internal magnetic field is strongly inhomogeneous and differs from He . It can be evaluated as [39, 52]:

200

100

0 0.0

0.1

0.2

0.3

0.4

0.5

z/w

Figure 26. The profile of the internal field in a stripe. H ∗ , indicating partial magnetic saturation of the stripe. For He > H ∗ , a broad band of nonresolved spin wave excitations is seen [60] in the low-frequency part of experimental spectra. At higher He , at first one and then more narrow peaks are observed in addition to the band. To understand the appearance of several multiple states in a SWW, one should take into account that the depth of the well strongly depends on He . In fact, the bottom of the well corresponds to Hi = 0 independently of He . The field, which determines the position of the top of the well, however, can be roughly estimated as He − Hd y = 0 ≈ He − H ∗ . For a

band

Hi = 600 Oe

Intensity (a.u.)

the micromagnetic system with interaction, in contrast to the standard single-particle situation [56]. These properties are not studied so far. It is, however, common practice to use a simple approximation describing the fluctuating field in the form of )-correlated random noise [55, 57, 58]. We believe that the results of simulations with such a deltacorrelated noise should provide at least qualitatively a correct picture of the spatial distribution of different magnetic eigenmodes in the system. In frame of this approximation the noise power has been calculated for a system of interacting magnetic cells as in [57]. The stochastic LLG equation has then been solved using the modified Bulirsch–Stoer method (details of the method are given elsewhere). The results of the simulation are presented in Figure 28. Figure 27 presents a scanning force microscope image of the rectangular elements studied and the distribution of the static magnetization obtained by solving the LLG equation without any fluctuations and used as a starting point for the dynamic simulation. Next the field Hifl t has been turned on. After the thermodynamic equilibrium state of the system has been reached, the obtained values of all cell moments were recorded. Afterwards, using Fourier analysis, the oscillation spectrum of the total magnetic moment of the element was obtained. In the frequency interval 3–20 GHz, this Fourier spectrum demonstrates several maxima. The Fourier components of each cell magnetization i " corresponding to these maxima were calculated and squared, to compare them with the measured spin wave intensities. As an example, the obtained spatial distributions for the frequencies  = 53 GHz and  = 122 GHz are presented in Figure 28. It is clear from Figure 28 that the low-frequency mode is strongly localized in the narrow regions near the edges of the elements that are perpendicular to the applied field. In contrast, and for comparison, the mode with  = 122 GHz shown in Figure 28 is not localized. Thus, we have shown that a strong inhomogeneity of the internal field causes the creation of “spin wave wells” (SWW) in magnetic stripes and rectangular elements and the localization of spin wave modes in these wells. It is clear from the above consideration that the number of the states localized in the well depends on the depth of the well, which is controlled by the value of the applied external field. This is demonstrated in Figure 29. Shown is a BLS spectrum for a transferred wavevector q = 047 · 105 cm−1 for a higher external field of He = 800 Oe instead of He = 500 Oe as it is the case for Figure 22. As it is seen, in addition to the PSSW modes, the spectrum contains two spin wave eigenstates of the SWW (instead of one in Fig. 22).

713

localized states

PSSW

q He

b)

4

6

8

10

12

Hi = 0

14

16

18

20

22

Frequency (GHz) H = 600 Oe

Figure 27. (a) AFM image of rectangular elements; (b) calculated distribution of static magnetization in an element.

Figure 29. BLS spectrum for a transferred wavevector q = 047 · 105 cm−1 and an external field of He = 800 Oe. PSSW indicates the perpendicular standing spin wave. The inset shows the used experimental geometry.

Dynamic Processes in Magnetic Nanostructures

714 22

800

Hi > 0

Internal field (Oe)

Frequency (GHz)

Hi = 0 14

10

600 4 400 ν = 6.2 GHz 2

band

200 yl

6

2

yr

0

H*

localized states

0.0

0.1

0.2

0.3

Wavevector (105 cm-1)

6

18

0.4

0 0.5

y (µm) 0

200

400

600

800

1000

External field (Oe)

Figure 30. Frequencies of the modes observed in the stripes at q = 047 · 105 cm−1 as a function of He . The vertical dashed line marks the critical field H ∗ . Note the constant frequency of one PSSW mode and the increase in frequency for the other mode for He > H ∗ . The gray region illustrates the band of nonresolved laterally quantized excitations.

small difference He − H ∗ the well is too shallow, and there is no room in the well for a localized spin wave state. With increasing He the well becomes deeper, and as a result of that, at first one, and then more localized spin wave states appear in the well. To obtain a complete quantization condition for localized spin wave states in a SWW, we need to modify Eq. (26) by adding phase jumps at the turning points of the SWW:  yr qHi y  dy = 2n (27)